Release Notes

Documentation

VoltDB Home » Documentation » Release Notes

Release Notes


Product

VoltDB

Version

7.7.8

Release Date

April 13, 2021

This document provides information about known issues and limitations to the current release of VoltDB. If you encounter any problems not listed below, please be sure to report them to support@voltdb.com. Thank you.

Upgrading From Older Versions

The process for upgrading from the recent versions of VoltDB is as follows:

  1. Shutdown the database, creating a final snapshot (using voltadmin shutdown --save).

  2. Upgrade the VoltDB software.

  3. Restart the database (using voltdb start).

For DR clusters, see the section on "Upgrading VoltDB Software" in the VoltDB Administrator's Guide for more special considerations related to DR upgrades.

Support for upgrading using shutdown --save was only added in V6.8. If you are upgrading from older versions of VoltDB, you will need to save and restore the snapshot manually. The procedure to do that is as follows:

  1. Place the database in admin mode (using voltadmin pause).

  2. Perform a manual snapshot of the database (using voltadmin save --blocking).

  3. Shutdown the database (using voltadmin shutdown).

  4. Upgrade the VoltDB software.

  5. Initialize a new database root directory (using the voltdb init --force action).

  6. Start the database in admin mode (using the voltdb start --pause action).

  7. Restore the snapshot created in Step #2 (using voltadmin restore).

  8. Return the database to normal operations (using voltadmin resume).

For customers upgrading from V6.x or earlier releases of VoltDB, please see the V6.0 Upgrade Notes.

For customers upgrading from V5.x or earlier releases of VoltDB, please see the V5.0 Upgrade Notes.

For customers upgrading from V4.x or earlier releases of VoltDB, please see the V4.0 Upgrade Notes.

Changes Since the Last Release

Users of previous versions of VoltDB should take note of the following changes that might impact their existing applications.

1. Release V7.7.8 (April 13, 2021)

1.1.

Ability to cancel a shutdown snapshot

The voltadmin shutdown --save command pauses the database, flushes buffers for services for export and database replication, and saves a final snapshot of the database contents. However, if remote systems are not available to acknowledge export or DR buffers, the shutdown may hang and you need to CTRL-C to exit from the command. When this happens, the shutdown process is still pending and the database not in normal operating mode. The voltadmin shutdown --cancel command has been added that lets you cancel a pending shutdown and resume normal operations.

1.2.

Recent improvements

The following limitations in previous versions have been resolved:

  • It was possible for busy clusters using command logging to end up generating back to back truncation snapshots, impacting performance. Safeguards have been put in place to avoid consecutive snapshots.

  • There was an issue where if a graceful shutdown operation was interrupted (for example, by a CTRL-C on the voltadmin shutdown command), a subsequent voltadmin shutdown --force command would fail. This issue has been resolved.

  • The snapshotconverter utility lets you generate CSV files from VoltDB snapshot files. These files can be used to recover and reload data from individual tables through the csvloader utility. However, for certain data — such as XDCR tables, tables defined with MIGRATE, or views with no COUNT(*) column — the snapshotconverter utility includes hidden columns in its output, which can be confusing. A new command flag has been added, --filter-hidden, that lets you exclude these hidden columns from the utility's output.

2. Release V7.7.7 (December 28, 2020)

2.1.

DR initialization snapshots changed to asynchronous processing

At the beginning of database replication (DR), a snapshot of the database is created and sent to the joining cluster. Previously, the initialization snapshot was created as a synchronous snapshot — blocking transactions on the existing database until initialization is complete. However, depending on the size of the database, the snapshot could take a significant amount of time to complete, stalling ongoing database transactions until the snapshot is complete.

This release changes the processing of DR initialization snapshots from synchronous to asynchronous. The asynchronous snapshot eliminates the interruption to ongoing work on the active cluster. The one drawback to this change is, when using cross datacenter replication (XDCR) with more than two clusters, if a node fails on the active cluster during the initialization snapshot, existing XDCR connections to other clusters may be lost and need to be reset.

3. Release V7.7.6 (July 2, 2020)

3.1.

SNMP improvements

All SNMP traps are now recorded in the log file. Previously, SNMP traps were sent, but not recorded. Log entries provide additional confirmation that SNMP events are triggered.

3.2.

Recent improvements

The following limitations in previous versions have been resolved:

  • There was an issue where SNMP traps would fail if the server had been running for around 24 days. After 24 days, whenever an SNMP trap was triggered, VoltDB would report that the trap failed with an illegal argument exception error. This issue has been resolved.

  • There was an issue where, if an XDCR cluster or a passive DR replica was started but not yet connected to another cluster, calling the @Statistics system procedure for DRPRODUCER or DRCONSUMER statistics would cause the statistics procedure to stall, to the point where the call would eventually time out. This issue has been resolved.

4. Release V7.7.5 (December 9, 2019)

4.1.

Recent improvement

The following limitation in previous versions has been resolved:

  • There was an issue with database replication where replication could stop and report that the replica was "ahead" of the master. This issue has been resolved.

5. Release V7.7.4 (October 1, 2019)

5.1.

Recent improvement

The following limitation in previous versions has been resolved:

  • If a node rejoining a cluster was stopped with @StopNode but later successfully rejoined the cluster, there was a chance the other nodes in the cluster did not recognize the successful rejoin and would continue reporting the error "REJOIN: No stream snapshot ack message was received in the past 10 minutes or the thread was interrupted." This issue is now resolved.

6. Release V7.7.3 (August 6, 2018)

6.1.

Recent improvements

The following limitations in previous versions have been resolved:

  • There was an issue with database replication (DR) where a large multi-partition transaction could produce binary logs from each partition on the producer that fell under the 50MB limit on inter-cluster communication. But when aggregated on the consumer for replay, the transaction exceeds the limit. The symptom was that nodes on the cluster would report a "bad message length" exception, causing nodes to be expelled from the cluster until the cluster itself failed. The possibility of an excessively large transaction still exists, but now the producer cluster rejects the transaction and replication and consistency between the clusters is maintained.

  • There is a condition where, under certain conditions, if the database is idle (that is, no read or write transactions are occurring) snapshots can get into a scheduling loop causing a CPU spike and preventing other threads from running. This occurs only when the database is configured with a large number of sites per host running on systems with slower disks and fewer CPU cores (for example, in virtualized environments). To avoid this condition, a new option, DISABLE_​IMMEDIATE_​SNAPSHOT_​RESCHEDULING, has been added.

    In normal database operation, this option is not needed. However, if you configuration matches these conditions and your database falls idle for any significant time, you can set this option to true when you start the database to circumvent the problem. You set the option as a Java environment variable on all the servers at start up using the VOLTDB_OPTS environment variable and including the "-D" flag. For example:

    export VOLTDB_OPTS="-DDISABLE_IMMEDIATE_SNAPSHOT_RESCHEDULING=true"
  • Due to a change in the server logic for the VoltDB Management Center (VMC) introduced in V7.7.2, the database replication (DR) and Import tabs no longer showed up when those features were turned on in the database. This issue has been resolved.

7. Release V7.7.2 (July 18, 2018)

7.1.

Reuse of HTTP connections to improve performance of JSON API and VMC

Previously, if security was enabled for a VoltDB database, JSON calls to the database (through the HTTP port) had to specify the username and password on each call. This could result in unnecessary and detrimental overhead as each call was authenticated. VoltDB now uses Keep-Alive, or persistent, HTTP connections. This means the connection is only authenticated once at the beginning of the session until the session ends or the connection becomes inactive (that is, no call is made) for the timeout period (10 seconds by default). Note that, even after a timeout, the session can be reestablished by passing the username and password again.

Use of Keep-Alive connections provides a significant performance improvement to both the JSON API and the web-based VoltDB Management Center, which uses the JSON interface. Note, however, for customer applications using the JSON interface, you must capture and return the session cookie that is provided by the JSON interface to take advantage of the performance improvement.

7.2.

Recent improvements

The following limitations in previous versions have been resolved:

  • There was a race condition that could cause a replica or XDCR cluster node to stop responding to calls to the @Statistics system procedure if an @Statistics call is received simultaneously to the node rejoining the cluster. This issue has been resolved.

  • There was a problem with the voltadmin status command, which worked as expected when used with the --continuous flag, but could fail with a traceback if run without the flag. This issue has been resolved.

8. Release V7.7.1 (November 9, 2017)

8.1.

Recent improvements

The following limitation in previous versions has been resolved:

  • There was a race condition introduced in VoltDB 7.7, where the voltadmin stop command could fail to stop the specified server, but not report any error. Worse yet, the race condition would cause the cluster to hang. The only way to free the cluster from the deadlock was to kill the server process on the server that failed to stop. This issue is specific to V7.7 and has now been resolved.

9. Release V7.7 (October 2, 2017)

9.1.

Support for multiple SQL statements in simple stored procedures

Simple stored procedures are stored procedures that can be declared as individual SQL statements without any further processing. They are declared using the CREATE PROCEDURE AS statement. You can now create simple procedures with multiple SQL statements by enclosing the statements in a BEGIN-END clause. For example:

CREATE PROCEDURE ReadyToShip 
   AS BEGIN 
     INSERT INTO shipments 
         SELECT order_num, item, customer FROM orders
         WHERE item=?;
     SELECT COUNT(*) FROM shipments;
   END;

See the description of the CREATE PROCEDURE AS statement in the Using VoltDB manual for details.

9.2.

Enhancements to the Analysis tab in VoltDB Management Center

The Analysis tab, which was added to the web-based VoltDB Management Center in 7.6, has been extended and now provides analysis of both procedure execution and table storage. Within the Analysis tab, after running the data gathering, you can choose between Data and Procedures to view statistics regarding the distribution of the tables and the execution of stored procedures.

9.3.

New and improved Go client

The VoltDB Go client has been revised to significantly improve performance, remove known defects and keep up to date with advances in the Go language. A description of the changes are available on the VoltDB website.

9.4.

New fraud detection example application

The VoltDB software kit includes a new sample application. The Fraud Detection example application demonstrates how easy it is to integrate Apache Kafka and VoltDB to create a fast data pipeline with data analysis and reporting. See the README in the examples/fraud-detection/ folder where you installed VoltDB for more information.

9.5.

Improved performance for distributed tasks within VoltDB

Recent enhancements to VoltDB improve the overall performance of certain distributed tasks. Specifically, concurrent read-only multi-partition stored procedures execute more rapidly due to better distribution of the procedures among the execution sites on each server.

9.6.

Enhanced voltadmin status command

The voltadmin utility has been enhanced to use a topology-aware client interface when providing ongoing cluster status. This means that when using the voltadmin status command to monitor a K-safe cluster, the voltadmin connection is maintained even if the database server to which the command was initially submitted fails.

9.7.

Deprecation notice

Over the next few releases, certain older, obsolete, or unused features of VoltDB will be deprecated. This means that the feature, although still operational, is no longer recommended for use and may be removed from the product in some future major release.

For example, recent performance work has minimized the speed difference between "safe" and "fast" consistency modes, if not eliminating it entirely. As a result, the "fast" option is being deprecated and will be removed in a future release. Note that this change requires no modifications to customer applications and is transparent to users.

9.8.

Security Notice

No security changes in this release.

9.9.

Additional improvements

In addition to the new features and capabilities described above, the following limitations in previous versions have been resolved:

  • The voltadmin update command incorrectly allowed changes to the sitesperhost setting when modifying the configuration of a running database. This issue has been resolved and the voltadmin update command now rejects changes to the sitesperhost setting.

  • If a cluster was started in admin mode (with the voltdb start --pause command), import connectors would not start as expected when the database was resumed, even if those connectors were enabled in the configuration. This issue has been resolved.

  • Similarly, if a K-safe cluster was started in admin mode and one of the nodes failed and rejoined before the database was resumed, the rejoining node would start in normal operational mode, not admin mode as it should have. This issue has been resolved.

  • The sqlcmd FILE -INLINEBATCH directive lets you define a sequence of DDL statements to process as a single batch, rather than one at a time. The directive works as described when included in a file. However, in previous releases, use of FILE -INLINEBATCH did not work when entered interactively at the sqlcmd prompt. This issue has been resolved and the directive works as documented for both file and interactive use.

  • VoltDB retains a cache of memory used to undo transactions. After each stored procedure finishes, the memory used for undo actions is held for the next transaction. In the past, that cache would grow to the largest size needed by any transaction. With this release, the size of the retained cache has been bounded to a much smaller size to avoid holding unused memory as a result of outlier transactions with unusually large undo requirements.

  • The --missing argument lets you start a VoltDB cluster with less than the full complement of nodes. However, VoltDB does not check that the resulting subset of nodes are sufficient to guarantee the cluster can start given the K-safety value. In previous releases, if it could not start, the server process would generate a null pointer exception. This issue has been resolved. VoltDB still does not guarantee the resulting subset of nodes can form a valid cluster, but now it simply does not start rather than issuing a misleading error.

  • In VoltDB 7.5, the kafkaloader utility was rewritten to use manual offset management, which improves reliability. However, the original implementation also impacted performance, reducing the potential throughput rate. This issue has been resolved and performance has returned to previous levels.

  • There was a problem with the SQL parser where a CREATE STREAM statement would fail if there was no space between the stream name and the following open parenthesis. For example, "CREATE STREAM mystream( ... ". This issue has been resolved.

  • The voltadmin dr drop command lets you remove a cluster from an existing cross datacenter replication (XDCR) relationship. However, previous versions were too strict in determining if a cluster could be dropped. Specifically, the command would not let you drop a cluster unless the current number of clusters matched how many clusters were connected when the current cluster joined. That meant a new cluster joining the relationship could disable the dr drop command. This issue has been resolved.

  • There was a race condition that could, on very rare occasions, be triggered by a schema change while a bulkloader (such as csvloader, jdbcloader, etc.) was running. The symptom of the race condition is that the bulkloader would report a hash mismatch and shut down the database. This issue has been resolved.

  • There was a similar race condition that could be triggered by two simultaneous schema changes. On occasion, this situation could trigger a fatal hash mismatch that would shut down the database. This issue has been resolved.

10. Release V7.6.1 (September 26, 2017)

The following issue has been corrected:

10.1.

Race condition interferes with schema updates and statistics reporting

Since VoltDB 7.1, there has been a race condition triggered by frequent schema changes (including updates to stored procedure classes or configuration options). The schema update could conflict with the gathering of database statistics, resulting in a number of different symptoms, including:

  • Subsequent schema updates taking longer to process.

  • Calls to the @Statistics system procedure timing out.

  • Attempts to invoke the VoltDB Management Center (VMC) failing with the error "could not establish connection to Database."

  • Nagios and other monitoring systems reporting similar failures.

This issue has been resolved.

11. Release V7.6 (August 28, 2017)

11.1.

User-defined functions

It is now possible to define your own functions for use in SQL queries and data manipulation statements. User-defined functions are written as Java methods, which are then declared using the CREATE FUNCTION statement:

CREATE FUNCTION season FROM METHOD myapp.calendar.DateToSeason;

See the chapter on "Creating Custom SQL Functions" in the VoltDB Guide to Performance and Customization for details.

11.2.

Data loading utilities stay connected

VoltDB has a number of utilities for loading data from external sources — CSV files, JDBC databases, and Kafka streams — as well as a generic bulk loader API for creating your own loading utility. Previously, if the loaders lost contact with all of the servers in the VoltDB cluster, the loading process would stop. Now, the utilities will wait, periodically retrying the connection, until the VoltDB cluster becomes accessible again.

This new behavior makes the loaders more resilient to failure, both of the external source and the VoltDB database. However, if you wish to revert to the previous behavior, you can add the --stopondisconnect argument when starting the csvloader, jdbcloader, or kafkaloader utility.

11.3.

Better mechanism for specifying delimiters in the file export configuration

It is possible to export files with different delimiters than the standard comma-separated delimiters using the file export connector. However, it has been difficult to use certain non-standard characters in the past due to the way the XML configuration option for the delimiters property was processed. This behavior has been changed to provide a simpler and more understandable set of rules.

Now, special and non-printing characters should be entered as Java string literals. For example, the new line character can be entered as "\n". Also any XML special characters, such as ampersand and greater-than, should be encoded as HTML entities. See the section on the "File Export Connector" in the Using VoltDB manual for details.

11.4.

Improved visibility of continuation lines in sqlcmd interface

The sqlcmd utility lets you continue SQL statements across multiple lines and the statement is not processed until you enter the closing semi-colon and press return. However, previously there was no indication that the utility was still waiting for more of the statement. Now the utility indents the command line prompt to show when it is still waiting for more input. For example:

$ sqlcmd
1> SELECT o.order_id, o.item, c.customer_id
  2> FROM orders as o,customers as c
  3> WHERE c.customer_id = o.customer_id
  4> AND o.ship_date > NOW();
5>

11.5.

Enhanced logging of schema changes

All changes to the database schema and/or the configuration are now recorded in the log file. The log records the time of the change, as well as the username and IP address of the person who changed it.

11.6.

New analysis tab in VoltDB Management Center

The web-based VoltDB Management Center has a new tab, Analysis, that lets you analyze the performance of a running database. The Analysis tab takes a snapshot of run-time database performance and provides several views of how the stored procedures are performing, including average execution time, frequency, and the total impact (that is, frequency times execution time. By clicking on a specific stored procedure, you can see further details on each statement within the procedure.

11.7.

Further improvements to the performance of schema updates

Over the past few releases, significant effort has been focused on reducing the performance impact of database modifications, such as schema changes, loading and unloading stored procedure classes, and modifying the database configuration. In particular, this release minimizes the amount of time transactions must wait while the changes are actually applied across the cluster.

11.8.

Security Notice

The following changes have been made to improve security and eliminate potential threats:

  • There were two HTTP endpoints that did not require authentication, even when security was enabled on the server, and could reveal the database schema. One endpoint has been removed and the other has been secured using standard VoltDB authentication mechanisms.

11.9.

Additional improvements

In addition to the new features and capabilities described above, the following limitations in previous versions have been resolved:

  • As a result of ongoing work to optimize schema changes, some internal procedure names have changed. For example, schema changes are no longer reported as an application catalog update, but rather as UpdateCore. These name changes do not impact user applications, but may change the information reported by system procedures such as @Statistics.

  • There was an issue where if a node failed in a K-safe cluster, it might not be able to rejoin the cluster, reporting that it "timed out waiting for the agreement site to recover." This problem was extremely rare, caused by a race condition when a link was lost between two nodes of a cluster only but the rest of the cluster connectivity was unaffected. This race condition has been corrected.

  • Previously, attempting to stop one node of a two node cluster (using @StopNode or the voltadmin stop command) might fail because it was erroneously identified as a potential network partition. It is now possible to stop either node of a two node cluster using these procedures.

  • There was an issue related to database replication (DR) where if a join inside a view references a DR table and data is restored to that view from a snapshot, the next attempt to modify that table would crash the database with the error "DR sequence number is out of sync with UniqueiId." This issue has been resolved.

  • Previously, if a database continually failed to write snapshots due to lack of disk space, ultimately the snapshot process would hang and no further snapshots were written, even if sufficient disk space was freed up. This issue has been resolved.

  • In VoltDB V7.5, attempting to perform an INSERT INTO SELECT where the target and destination tables are the same table, would fail, causing VoltDB to use up all available system memory and either hang the system or crash the database. This issue has been resolved.

  • Starting with V7.5, it was possible for an upgrade from previous versions of VoltDB to fail during the snapshot restore due to a DDL error, such as "object name already exists". This happened only if the schema had been previously updated using LOAD CLASSES to load a JAR file containing additional files beyond just the stored procedure class files. Specifically, any stray SQL files in the JAR were being improperly loaded as part of the restore operation. This issue has been resolved.

  • The VoltDB DECIMAL datatype is a 16-byte fixed scale decimal value. Previously, the csvloader utility did not properly handle strings representing values that exceeded this limit in terms of scale or precision. Instead, it threw an exception, resulting in the error "failed to drain all buffers, some tuples may not be inserted yet." This issue has been resolved.

  • Previously, attempts to write a small (63 bytes or less) VARBINARY value into a larger (64 bytes or more) VARCHAR column could crash the database due to an access violation or segmentation fault. For example, performing an INSERT INTO SELECT where the target column is defined as VARCHAR(64 BYTES) and the source column is VARBINARY(62). This issue has been resolved.

  • There was an issue with the export connector for file export in previous releases where, if file export failed (for example, due to a temporary a disk error) subsequent export would continue writing to the same output locations and the output files would no longer be "rolled" as expected. Further, if the current output files were deleted, export data would no longer be written. This issue has been resolved.

12. Release V7.5 (July 28, 2017)

12.1.

Support for Java boxed datatypes

Java boxed datatypes such as Short, Integer, and Long are now supported as input to stored procedures. Previously, only primitive datatypes were supported. Using boxed types, you can pass Java null values directly to VoltDB and have them interpreted correctly.

12.2.

New functions to convert between radians and degrees

Two new SQL functions have been added, RADIANS() and DEGREES(), that let you convert floating-point values between radians and degrees.

12.3.

Improvements to offset tracking for the kafkaloader utility and Kafka import

The kafkaloader utility and built-in Kafka importer have been modified to improve the tracking of the current offset between Kafka and VoltDB. By default, these features now use a manual commit policy to ensure the current offset in the Kafka message queue matches the records successfully inserted into the VoltDB database, rather than tracking just what messages have been successfully read from Kafka. You can, if you wish, revert to an automatic commit policy by setting the appropriate property in the command line argument or import configuration.

12.4.

Support for SSL and Kafka 0.10.2

Kafka 0.10.2 supports secure socket layer (SSL) encryption. Using the 0.10.2 version of the kafkaloader utility (kafkaloader10), you can connect to SSL-secured Kafka brokers by putting the appropriate SSL properties in the Kafka configuration file specified in the --config argument. For example, using a local trust and key store, the properties file might look like this following. (This only works with the kafkaloader10 utility, since Kafka version 0.8.2 did not support SSL.)

security.protocol=SSL
ssl.truststore.location=/Home/jdoe/mytruststore.jks
ssl.truststore.password=mytrustpasswd
ssl.keystore.location=/Home/jdoe/mykeystore.jks
ssl.keystore.password=mykeypasswd
ssl.key.password=passwd

12.5.

Ability to write Kafka messages to two or more tables

The Kafka import connector is now capable of writing messages from a single topic to multiple tables, without requiring a custom insert procedure. If you create two configurations with the same Kafka topic ID but different group IDs, you can have a single topic written to two different tables. For example, the following configuration writes records from the Kafka employees topic to both the EMP and MGR tables.

<import>
   <configuration type="kafka" enabled="true">
       <property name="topics">employees</property>
       <property name="groupid">voltdbemp</property>
       <property name="procedure">EMP.insert</property>
   </configuration>
   <configuration type="kafka" enabled="true">
       <property name="topics">employees</property>
       <property name="groupid">voltdbmgr</property>
       <property name="table">MGR.insert</property>
   </configuration>
</import>

12.6.

New @Statistics selector QUEUE

A new selector has been added to the @Statistics system procedure. The QUEUE selector returns statistics on the number of tasks in each partition's process queue and the average and maximum time tasks were waiting in the queue. See the description of the @Statistics system procedure in the Using VoltDB manual for details.

12.7.

Improvements to INSERT INTO... SELECT to reduce temporary memory usage

Performing an INSERT INTO... SELECT can produce many records, depending on the results of the SELECT. In the past, it was easy to use up all of the available temporary memory that VoltDB allocates for transactions, resulting in the statement failing and rolling back. The execution plans for these statements have been rewritten to perform each insert inline, significantly reducing the amount of temporary memory required and increasing the number of rows that can be processed in a single statement.

12.8.

More flexibility in placement of COUNT(*) in CREATE VIEW

Previously, the CREATE VIEW statement selection list had to begin with any columns listed in the GROUP BY statement, followed by COUNT(*), and then any other aggregate functions. The restriction on the placement of COUNT(*) has been removed and it can now appear anywhere in the selection list after the GROUP BY columns.

12.9.

Ability to use absolute paths to identify custom import formatters

By default, when using a custom import formatter, VoltDB looks for the formatter JAR file in the bundles directory where VoltDB is installed. However, you can now use a full path URL (starting with "file:") to specify any arbitrary location for the JAR file. See the description of installing and invoking custom importers in the VoltDB Guide to Performance and Customization manual for details.

12.10.

Dynamic connections to clusters for the kafkaloader and jdbcloader utilities

The kafkaloader and jdbcloader utilities have been updated to use the auto-reconnecting client library, so they can adjust to changes in the database cluster configuration dynamically.

12.11.

New message logged when processes run more than 10 seconds

If a task blocks a process queue for more than 10 seconds, an informational message is written to the server log. You can adjust the threshold at which these messages are written by setting the configuration element <procedure loginfo="{milliseconds}"/> in the <systemsettings> section of the cluster configuration file. See the appendix on server configuration options in the VoltDB Administrator's Guide for details.

12.12.

Improved partition management when cluster nodes fail

When a node fails on a K-safe cluster, management of one or more of the database partitions (known as mastership) may need to be re-assigned to the remaining nodes. In previous versions, after several failures and rejoins mastership of the partitions could become unevenly dispersed in the cluster. Reassignment has now been improved to maintain an even distribution in the cluster.

12.13.

Additional improvements

In addition to the new features and capabilities described above, the following limitations in previous versions have been resolved:

  • Previously, VoltDB did not support what is called unary minus. That is, where a column name or other reference has a minus sign in front of it. For example SELECT item, in_stock, -ordered from PRODUCT. Unary minus is now supported.

  • The VoltDB Java client includes a timeout value that is user settable and breaks a hung connection after the specified time limit. However, if the server went down while the Java client was initially authenticating the connection, the timeout was not observed and the client process would hang indefinitely. This issue has been resolved and authentication requests now obey the connection timeout limit.

  • Previously, there was an edge case where, if a SQL statement utilizes certain indexes, evaluates a COUNT(*) function, and attempts to compare a VARCHAR column value to a string longer than the column's maximum length, the statement would result in a runtime error. For example, under these conditions if PRODUCTCODE is defined as VARCHAR(2), the expression "WHERE PRODUCTCODE = 'ABC'" would return an error rather than false. This issue has been resolved.

  • Previously, updating the stored procedure classes (using the sqlcmd LOAD CLASSES directive or the @UpdateClasses system procedure) updated all of the classes in the schema. As a result, the more classes in the schema, the longer the update would take, even if the update itself was small. VoltDB now updates only those classes affected by the change, significantly reducing the time required for the update to complete.

  • Previously, specifying a schema or a schema and a JAR file of Java classes using the --schema and --classes arguments to voltdb init would successfully load the specified items on start. However, specifying just a JAR file did not result in the classes being loaded on start. This issue has been resolved and use of --classes without --schema now works.

  • There was also a race condition in the BulkLoader, aggravated by frequent insert failures, where it could lose track of exactly how many failures had occurred. The consequences of this race condition were that the loader utilities (such as csvloader) could report an incorrect number of failures and for applications using the BulkLoader API, any callback associated with failures might not be invoked for all failure cases. This issue has been resolved.

  • Previously, if a stream column was declared as NOT NULL in the CREATE STREAM statement, it was still possible to insert a null value for that column into the stream. This issue has been resolved and the NOT NULL constraint is now properly enforced at runtime.

  • Normally, dropping a table is not possible if a stored procedure exists that references that table. However, in the past if a stored procedure only accessed the table by a COUNT(*) aggregate (for example, SELECT COUNT(*) from MyTable) the dependency was not properly recognized and the DROP TABLE could succeed. Subsequent calls to the procedure would then generate errors. This issue has been resolved and such procedures now correctly block any DROP TABLE statement for the referenced table.

  • An issue was introduced in 7.4 where, if the database was configured to import data and the configuration or schema related to import was modified while the database was paused (or a node rejoined the cluster while the database was paused), the database could crash. This issue has been resolved.

13. Release V7.4 (June 19, 2017)

13.1.

New DR DROP command

There is a new voltadmin command to safely remove a single cluster from a multi-cluster XDCR environment without disrupting the connection between the remaining clusters. The DR DROP command tells a running cluster to send any queued binary logs, stop DR and inform the remaining clusters to remove it from the XDCR environment. See the description of the voltadmin utility in the Using VoltDB manual for details and an explanation of the difference between DR DROP and DR RESET.

13.2.

Support for alternate character sets in the csvloader utility

The csvloader utility has a new option, --charset, that lets you specify the character set of the input file. See the description of the csvloader utility in the Using VoltDB manual for details.

13.3.

New column added to the @Statistics DRPRODUCER selector output

A new column, QUEUE_GAP, has been added to the return values of the DRPRODUCER selector of the @Statistics system procedure. This column indicates if there is a gap between the last acknowledged transaction and the next transaction queued for transmission to the DR consumer. See the description of the @Statistics system procedure in the Using VoltDB manual for details.

13.4.

The sqlcmd FILE directive now supports processing multiple files in a single batch

Previously, the sqlcmd FILE directive let you process a single file containing SQL statements (and other sqlcmd directives). The directive has been extended to support a list of files, separated by spaces. If you include the -batch argument, all of the files are processed as a single batch. See the description of the sqlcmd utility in the Using VoltDB manual for details.

13.5.

Support for converting Java BigDecimal to integer VoltDB datatypes

Java BigDecimal values are now allowable input to VoltDB integer datatypes, including TINYINT, SHORTINT, INTEGER, and BIGINT. As with any datatype conversion, if the BigDecimal value overflows the target integer type it will generate a runtime error. See the appendix on datatype compatibility in the Using VoltDB manual for details.

13.6.

The voltdb init --schema option now ignores unsupported sqlcmd directives

The --schema option to the voltdb init command lets you load a preliminary schema when initializing the database root. However, originally the schema file could only contain SQL DML statements. So certain files that were acceptable to sqlcmd were not acceptable to the voltdb init command because the files contained sqlcmd directives. The voltdb init command now ignores these directives and successfully processes the accompanying DML statements.

13.7.

Optimization of runtime schema changes

As part of an ongoing effort to optimize and reduce the impact of changes to the running database — specifically changes to the schema, stored procedures, or configuration — the internal code to implement these actions is being reorganized. One consequence of this change is that the name of internal procedures reported by the @Statistics system procedure may change. For example, previously, schema changes were reported by the @Statistics INITIATOR and PROCEDURE selectors as the UpdateApplicationCatalog procedure. Now these actions are executed by the UpdateCore procedure.

13.8.

Hash indexes are now deprecated

VoltDB uses tree indexes by default. It has been possible to request a hash index by including the string "hash" in the index name. However, testing shows that in most cases, tree indexes are more effective or at least as effective as hash indexes. Consequently, hash indexes are now deprecated and will be removed at an upcoming major release.

13.9.

Additional improvements

In addition to the new features and capabilities described above, the following limitations in previous versions have been resolved:

  • There was a very rare condition where a hash collision in the planner could cause it to select the wrong SQL plan on one node. Although still extremely rare, the chances of a collision were increased with frequent schema and procedure changes. If this happened, the database would detect a hash mismatch on the transaction, shutting down the database to preserve data integrity. This issue has been resolved.

  • In cross datacenter database replication (XDCR), if the schema changes on one cluster, DR interaction should stop until the other cluster(s) are updated to match. However, previously a DROP TABLE statement on a DR table was not correctly recognized as a schema change and DR was not paused. This issue has been resolved.

  • There was an issue with database replication (DR), where a producer node compared the last acknowledged ID to the last queued ID to see if DR was completely drained. However, it is possible for a server that was offline temporarily to not have some of the DR logs that the consumer needs, which another unavailable server does have. However, the producer would erroneously report that the queue was successfully drained. This issue has been resolved and the clusters now takes gaps in the queue into consideration when reporting if the DR is drained or not in response to a voltadmin pause --wait or shutdown --save command. If missing logs exist but are not currently available (because a server is down), the commands will report this to the user.

  • Previously, there was an issue when using the Kinesis importer and defining two configurations that read from the same Kinesis stream but using different import procedures. The symptom was that the server reported frequent class not found exceptions. This issue has been resolved.

  • When using the file export connector, the delimiters property is supposed to let you change the delimiters for the CSV output. However, the string value was not being properly interpreted. This issue has been resolved.

    Note, however, that the documentation was also not complete in its description of how the property was meant to be interpreted. The delimiters property value must obey the following rules:

    • The string represents four delimiter characters: the separator, the quote character, the escape character, and the end-of-line character.

    • All four delimiters must be specified in the aforementioned order.

    • Special and non-printing characters must be encoded as HTML entities for inclusion in the XML configuration file. For example encoding "<" as "&gt;" and new line as "&#10;".

    • The new line character is the only white space character that can be used as the end-of-line delimiter. Space, tab, carriage return and so on cannot be used in that position.

  • When a VoltDB server process is starting up, especially during a rejoin, there was a narrow window when the programming interface became available but the database replication state was not fully instantiated. It was possible for a call to the @Statistics DRROLE selector during that window to stall. This issue has been resolved.

  • An issue was introduced in VoltDB 7.2, where if the stored procedure classes are updated without any accompanying changes to the actual schema, a cache of ad hoc statements that should have been updated was not. As a result, subsequent ad hoc queries using that cache would hang and block any further ad hoc queries. This issue has been resolved.

14. Release V7.3 (May 24, 2017)

14.1.

New online upgrade process for VoltDB software

There is a new process available for upgrading VoltDB software on a running cluster using XDCR without requiring a second set of cluster hardware. Upgrading the software between two XDCR clusters is the easiest method. However, if you do not want or have the resources for a second separate cluster, the new process uses XDCR and K-safety to split an existing cluster into two and upgrade the two halves through XDCR. See the section on "Performing an Online Upgrade with Limited Hardware" in the VoltDB Administrator's Guide for details.

14.2.

New @Statistics selector for measuring latency

There is a new selector for the @Statistics system procedure, LATENCY, which provides periodic latency statistics. The statistics are generated every five seconds and provide both maximum and various levels of percentile measurement up to 99.999%. See the description of the @Statistics system procedure in the Using VoltDB manual for details.

14.3.

Support for Kerberos when using the VoltDB command line utilities

Kerberos support has been extended to provide integration with the VoltDB command line utilities, such as sqlcmd and voltadmin. See the discussion of "Accessing the Database from the Command Line" in the Kerberos section of the Using VoltDB manual for details.

14.4.

Beta release of kafkaloader utility support for Kafka version 0.10.2

There is a new version of the kafkaloader utility, named kafkaloader10, that works with Kafka version 0.10.2. There are enough differences between Kafka 0.8.2 and 0.10.2 that two separate utilities will be maintained for the foreseeable future.

Customers interested in using kafkaloader10 are encouraged to try this beta release and provide us feedback. Note however, it has not yet received the rigorous testing of the default 0.8.2 version of kafkaloader and should not be considered production-ready quite yet. See the online help for kafkaloader10 in the bin directory for more information.

14.5.

Ability to choose a preferred source for a new XDCR connection

It is now possible to select one cluster in a multi-cluster XDCR configuration as the preferred host of the initial synchronizing snapshot when the current cluster joins the environment. You use the preferred-source attribute in the <connection> configuration to identify the cluster ID of the cluster that will send the snapshot. For example, the following configuration specifies that cluster ID 2 will send the snapshot when cluster ID 3 first joins the XDCR environment:

<dr id="3" role="xdcr" >
   <connection source="svra,svrb,svrc" preferred-source="2" />
</dr>

Note that preferred-source only affects the initial snapshot. All other DR communication occurs among all clusters within the XDCR environment.

14.6.

New Import tab in the VoltDB Management Center

There is a new tab available in the VoltDB Management Center that helps you monitor the progress of import connectors. The new Import tab is only visible when import connectors are configured and active.

14.7.

Improved performance and reduced bandwidth use in the VoltDB Management Center

The VoltDB Management Center (VMC) has been optimized to reduce its bandwidth use and thereby improve overall performance.

14.8.

Ability to export table and column names in lowercase for JDBC targets

By default, VoltDB sends SQL statements through the JDBC export connector using all uppercase table and column names. However, a few databases (most notably, PostgreSQL) are case-sensitive. There is a new property for the JDBC connector that tells VoltDB to use lowercase table and column names. The lowercase property accepts either true or false, with false being the default. See the description of the JDBC export connector in the Using VoltDB manual for details.

14.9.

Deprecation notice: Java client API method updateApplicationCatalog()

There is a method in the VoltDB Java client API, updateApplicationCatalog(), designed for internal use based on features that are no longer available in the product. This method is now deprecated and be removed from the Java client library in the next major release of VoltDB.

14.10.

Additional improvements

In addition to the new features and capabilities described above, the following limitations in previous versions have been resolved:

  • It was possible for the return values of the @Statistics PLANNER selector to overflow their INTEGER datatype. This issue has been resolved by increasing the size of the cache hits and misses columns to BIGINT.

  • There was an edge case where attempting to drop a table or stream from the database could cause a null pointer exception. This issue has been resolved.

  • The new online upgrade process using limited hardware allows you to upgrade from VoltDB versions starting with 7.2. However, in its initial release, the process did not support upgrading when SSL is enabled for the external ports. This issue has now been resolved. So it will be possible to perform online upgrades of SSL-enabled clusters from version 7.3 and later.

  • It was possible for the kafkaloader utility to encounter an unexpected error, stopping one of its threads, without notifying the user, giving the impression the load process completed successfully. This issue has been resolved and now unexpected errors cause the utility to stop, reporting the error to the user.

  • In certain circumstances, the Kafka export connector could drop a few rows between a database pause and resume. This issue has been resolved.

  • With security enabled, it was possible for VoltDB to report multiple occurrences of an error indicating it failed to read a message length while attempting to authenticate and register a user. There was no real underlying problem; the errors were caused by processes connecting and disconnecting without sending any data. This issue has been resolved and the spurious messages removed.

  • There was an issue where, if an XDCR cluster recovered while another XDCR cluster was also stopped, the first cluster could not properly reset replication on the downed cluster. The voltadmin dr reset --cluster command would not report an error, but the DR connection would not be reset. This issue has been resolved.

  • VoltDB 7.1 introduced improvements to the UPDATE CLASSES statement, that significantly reduced the time needed to load new or updated classes. Unfortunately, one of these improvements introduced a new issue where updated classes with new or modified SQL statements could result in incompatibilities and cause run-time errors. This modification has been backed out. Performance of UPDATE CLASSES is still better than versions prior to 7.1. However, the new change reduces the performance increase seen in 7.1 and 7.2.

  • VoltDB does not permit schema changes while a node is rejoining the cluster or a new node is being added elastically. However, there was a race condition where, if a rejoin and a schema change were initiated at approximately the same time, it was possible for both to occur simultaneously. The consequence was that the rejoining node would receive the old rather than the updated schema. Once the rejoin operation was complete, the different schema would result in a hash mismatch error, stopping the database. This rare condition has been corrected.

15. Release V7.2 (April 25, 2017)

15.1.

Ability to initialize root directory with schema and procedure classes

It is now possible to add a starting schema and procedure classes when you initialize the database root directory. The new options --schema and --classes let you specify a file containing DDL commands and a JAR file of stored procedure classes, respectively. These items are loaded automatically the first time the database is started. (And subsequent starts, if command logging is not enabled.) For example, the following command pre-loads DDL commands from the file myschema.sql and a set of procedure classes from myprocs.jar:

$ voltdb init --config=deployment.xml \
              --schema=myschema.sql --classes=myprocs.jar

See the description of the voltdb init command in the Using VoltDB manual for details.

15.2.

Enhancements to the DR reset command for managing XDCR environments

In multi-cluster XDCR environments, it is now possible to drop one cluster from the network without disrupting replication between the remaining clusters by using the voltadmin dr reset --cluster command. See the description of the voltadmin command in the Using VoltDB manual for details.

15.3.

DR support for @SwapTables

Use of the @SwapTables system procedure is now supported by database replication (DR). Swapping tables on a producer cluster acts like a schema change, causing the consumer cluster to pause until you execute the same @SwapTables statement on the consumer. See the description of the @SwapTables system procedure in the Using VoltDB manual for details.

15.4.

Support for subqueries in data manipulation language (DML) statements

Data manipulation language (DML) statements, such as INSERT, UPDATE, and DELETE, now support subqueries. See the description of the individual statements in the appendix of supported SQL statements in the Using VoltDB manual for details.

15.5.

New voltadmin status command

There is a new command, voltadmin status, that reports the current status of a running cluster. Information in the output includes the number and IP addresses of the servers in the cluster and if any nodes are missing from a K-safe cluster. You can also add the --dr option to include information on the current state of data replication for the cluster. See the description of the voltadmin command in the Using VoltDB manual for details.

15.6.

New voltdb collect syntax

The syntax for the voltdb collect command has been enhanced to be consistent with the other voltdb commands, such as get, init, and start. The voltdb collect command now let's you specify the location of the root directory using the --dir option, like other voltdb commands. It also has an --output option that lets you specify both the name and location of the output file. Note that, although deprecated, use of the old syntax will continue to be supported during the 7.x release cycle.

15.7.

Improved logging of security events and configuration changes

VoltDB now provides more thorough logging of security events. Specifically, every authorization failure is logged, each user who successfully authenticates is logged, and every configuration change is logged prior to the change being made.

15.8.

New documentation of custom exporters, importers, and formatters

A new chapter on writing custom exporters, importers and formatters has been added to the VoltDB Guide to Performance and Customization.

15.9.

Additional improvements

In addition to the new features and capabilities described above, the following limitations in previous versions have been resolved:

  • There was an issue where the BulkLoader programming interface could lose track of the correct number of processed records, resulting in a negative value for outstanding rows, which in turn caused the drain() operation to never complete. This issue has been resolved.

  • Previously, the JSON API returned the wrong error status for invalid URLs and lack of permission. It now returns error codes 404 and 401, respectively.

  • Previously, it was possible to define an index with an expression that could, at run-time, generate an error (for example, using a zero column value as the denominator of an expression). When offending data was inserted into the table, an error was correctly reported but the insert was still performed. Consequently, performing SELECT operations ordered by the erroneous index would not return all valid rows. VoltDB no longer accepts such index definitions on non-empty tables. As a result, some index definitions that were allowed in previous versions will not be possible in 7.2 and later.

  • Performing a port scan on the VoltDB internal port (port 3021 by default) previously resulted in the socket not closing properly. This issue has been resolved.

  • There was a small memory leak possible when a database replication (DR) producer cluster recovered using command logs. This issue has been resolved.

  • Previously, enabling SSL caused the VoltDB Nagios scripts to fail. This issue has been resolved.

  • Previously, when using cross datacenter replication (XDCR), if a node from a cluster involved in an XDCR environment was removed from the cluster and then used as a member of another cluster in the same XDCR environment, DR connections for that cluster would not start properly. This issue has been resolved.

  • The HTTP export connector would fail with a null pointer exception if any rows written to that connector included a VARCHAR column with a null value. This issue has been resolved.

  • Previously, if an XDCR cluster performed a voltadmin dr reset (which reset DR connections to all other existing clusters), adding a new cluster to the XDCR environment would not work properly. The original cluster would produce binary logs for the new cluster, but not consume the new cluster's logs. This issue has been resolved.

  • There was an issue where, although an index using the "IS NOT DISTINCT FROM" clause could be parsed and added to the database schema, the index definition was not properly re-generated on subsequent schema changes. The result was any further schema changes would fail with syntax errors. This issue has been resolved.

  • Until now, the Kafka importer only accepted incoming data as text. It now accepts binary data as well.

16. Release V7.1 (March 21, 2017)

16.1.

Support for TLS encryption on external ports

VoltDB now supports Transport Layer Security (TLS) — the recommended upgrade from Secure Socket Layer (SSL) encryption — for all externally accessible ports, not just the HTTP port. With TLS enabled, data passing through both the client and admin ports is encrypted. If encryption is enabled for the client and admin ports, you must use the --ssl option on the VoltDB command line tools (such as sqlcmd, voltadmin, and voltdb) to connect to the database. The Java and C++ client libraries also now support connecting through TLS-enabled ports.

Important

To support the new options, the configuration of TLS in the deployment file has been reorganized. Customers who previously enabled TLS encryption for the HTTP port only will need to change their configuration files. For example, the old and new syntax for enabling TLS on the HTTP port is:

Old Syntax

<httpd>
   <https enabled="true">
      [ . . . ]
   </https>
</httpd>

New Syntax

<ssl enabled="true">
   [ . . . ]
</ssl>

See the section on using TLS/SSL encryption in the VoltDB Administrator's Guide for details on setting up TLS for a VoltDB cluster.

16.2.

Ability to retrieve the schema, procedures, and configuration from a database

Three new voltdb commands have been added that allow you to extract the schema, stored procedures, and configuration from the local database root directory. These commands can be used whether the database is running or not. The new commands are:

voltdb get schema
voltdb get classes
voltdb get deployment

See the description of the voltdb command in the Using VoltDB manual for details.

16.3.

New statistics provide performance details about individual statements within a procedure.

A new selector, PROCEDUREDETAIL, for the @Statistics system procedure provides detailed performance information about the execution of individual statements within stored procedures. See the description of the @Statistics system procedure in the Using VoltDB manual for details.

16.4.

New system procedure lets you swap two tables

A new system procedure, @SwapTables, lets you swap the contents of two tables. The two tables must have the identical schema (that is, the same column names, datatypes, and constraints) and cannot have views defined. See the description of the @SwapTables system procedure in the Using VoltDB manual for details.

16.5.

Ability to restart a partial cluster

It is now possible to restart a K-safe cluster with fewer than the original number of nodes. For example, if one server fails on a five node cluster and you want to restart the cluster but the failed server is not ready, you can start the cluster with the remaining four nodes by using the --missing argument on the voltdb start command, like so:

$ voltdb start --count=5 --missing=1

Note that the partial cluster is not fully K-safe to its specified K-safety value until the full complement of nodes rejoin the cluster. See the description of the voltdb command in the Using VoltDB manual for details.

16.6.

New SQL functions for handling internet addresses

There are four new SQL functions to convert internet addresses from text strings to binary and back again. Two functions (INET_ATON and INET_NTOA) convert IPv4 addresses and two (INET6_ATON and INET6_NTOA) convert IPv6 addresses. See the Appendix of SQL Functions in the Using VoltDB manual for details.

16.7.

Statistics related to schema changes are no longer reset by each change

Previously, the information gathered by the @Statistics system procedure related to schema changes (from DML statements or calls to @UpdateClasses or @UpdateApplicationCatalog) was reset each time the database schema changed. Schema changes no longer reset the statistics on each invocation. Also, every schema change is recorded, rather than the 5% sampling done for other procedures.

16.8.

SNMP traps now report database replication (DR) events

VoltDB can send SNMP traps for selected database events, such as node failures or when resource limits are reached. Starting with V7.1, it also reports database replication (DR) events when DR is enabled.

16.9.

New index optimizations

We continue to extend support for the use of indexes to optimize the execution of SQL statements. Two new uses of indexes are when evaluating:

  • Window functions

  • NOT DISTINCT predicates

16.10.

Enhancements to the kafkaloader utility

Three new features have been added to the kafkaloader utility:

  • Support for a Kafka consumer configuration file, including the ability to set the group.id property. Use the --config argument to specify the configuration file.

  • Support for custom formatters. Use the --formatter argument to specify custom formatter properties.

  • Support for large partition counts. Use the -k argument to specify the number of Kafka partitions per topic if there are more than 10 partitions.

See the online help for the kafkaloader utility for details.

16.11.

Improved performance when updating procedure classes

Previously, updating the stored procedure classes (using the sqlcmd LOAD CLASSES directive or the @UpdateClasses system procedure) updated all of the classes in the schema. As a result, the more classes in the schema, the longer the update would take, even if the update itself was small. VoltDB now updates only those classes affected by the change, significantly reducing the time required for the update to complete.

16.12.

Managing heterogeneous clusters in Database Replication (DR)

You can create a DR environment using clusters running different versions of VoltDB. However, to support new functionality (in particular, multi-cluster XDCR) the DR protocol changed in V7.0. That major version acts as a gateway between earlier and later releases. In other words, you can run a V7.0 cluster with earlier versions, or you can run a 7.0 cluster with later versions. However, you cannot run pre-V7.0 clusters with post-V7.0 clusters.

As a consequence, if you want to perform an in-service upgrade (using different version clusters) to upgrade a DR environment using a pre-V7.0 version (say, V6.9) to V7.1 or later, you must upgrade all clusters to V7.0 first, then upgrade to your final target version. Of course, you can always use the normal single-cluster upgrade process to go directly from V6.x to V7.x if you choose.

16.13.

VoltDB servers no longer support SHA-1

To ensure the security of VoltDB servers, the database no longer accepts SHA-1 for hashed data. All hashed data (such as passwords etc) must be hashed using SHA-256. Note: this is an incompatible change. Some earlier client libraries will not be able to connect to the V7.1 servers because they do not support the more robust SHA-256 hashing.

16.14.

Additional improvements

In addition to the new features and capabilities described above, the following limitations in previous versions have been resolved:

  • There was an issue where, if multiple VoltDB client connections were created in separate threads of a Java application, a race condition could cause the clients to attempt to use shared resources before they were properly initialized. The symptom of this problem was that the connections failed reporting an illegal state exception accessing the ReverseDNSCache. This issue has been resolved.

  • Previously, VoltDB created a file (log/volt.log) in the user's current working directory whenever you started the sqlcmd utility. However, no messages were ever written to the log, so this superfluous file is no longer created.

  • If a database is run without command logging but with automated snapshots, the database is not restored when the database process is stopped and restarted with the voltdb start command. Automated snapshots are intended as a form of backup, not persistent durability. However, previously, backups from a previous session would eventually be deleted as part of the cleanup of older backups. To avoid this, when VoltDB starts without command logs, any snapshots present in the snapshot directory are archived to a separate directory so they will not be accidentally deleted.

  • If a database is not using command logging, each time you stop and restart it with the voltdb start command, it should start a fresh, empty database (with no durability). However, previously, if the earlier session used certain features (such as export) and left overflow data in the database root directory, the voltdb start command would not allow this data to be overwritten and would force the user to re-initialize the root directory. This problem has been corrected and the voltdb start command starts a fresh database instance, even if there is existing overflow data as long as no other durability data (such as command logs) is present.

  • In earlier releases, it was possible for the database server to report a ZooKeeper CancelledKeyException error during shutdown. This error was harmless, but unnecessary. It has been removed.

  • Previously, attempting to create a stream with 256 columns could crash the database. This issue has been resolved.

  • If a table contained a very large VARCHAR column and the table had a view defined including either a MIN or MAX aggregate function, deleting a row with a large value in the VARCHAR column could result in a fatal error. This issue has been resolved. It is still possible for the DELETE operation to fail, but now the appropriate error is returned to the user.

  • There was an edge case, where if a table had a VARCHAR or VARBINARY column that was indexed as part of a complex index and the column was also used in an MIN or MAX aggregate function of a view, updating that column could result in a fatal SIGSEGV error. This issue has been resolved.

  • Previously, changing the Log4j configuration on a running database (for example, by calling @UpdateLogging) and specifying a relative path for the log file, would result in the file location being interpreted relative to the user's current working directory. This has been corrected and relative paths are now interpreted as relative to the database root directory.

17. Release V7.0.1 (March 16, 2017)

17.1.

Recent improvements

The following limitations in previous versions have been resolved:

  • There was an issue in the voltadmin shutdown --save command introduced in 6.8. If command logging was not enabled on the cluster, the final snapshot would not be properly restored on the subsequent voltdb start. This problem has now been fixed.

  • There was an issue, introduced in V6.7, that caused database replication (DR) between mixed version clusters to fail when replicating from a 6.7 or later cluster to a pre-6.7 cluster. If the more recent version cluster's schema includes a view on a DR table, replication would fail to start or break when a schema change adds such a view. This issue has been resolved.

  • Previously, if a cluster was configured for cross datacenter replication (XDCR), but the deployment did not specify a <connection> element or the <connection> element was explicitly disabled (with the enabled="false" attribute), the cluster could generate a null pointer exception. This issue has been resolved and the error is no longer generated. Note, however, that explicitly disabling the connection is not a common configuration and XDCR replication will not operate properly until the connection is enabled.

  • The import infrastructure has been redesigned to reduce resource contention and optimize the way back pressure is handled. These changes should result in significantly improved import performance.

  • Previously, the @Statistitcs system procedure DRPRODUCER selector could erroneously report that there was data still waiting to be drained even though the queue was empty. This issue has been resolved.

18. Release V7.0 (January 30, 2017)

18.1.

Updated operating system and software requirements

The operating system and software requirements for VoltDB have been updated based on changes to the supported versions of the underlying technologies. Specifically:

  • Ubuntu 12.04 is no longer supported. The supported operating system versions are CentOS 6.6, CentOS 7.0, RHEL 6.6, RHEL 7.0 and Ubuntu 14.04 and 16.04, with support for OS X 10.8 and later as a development platform.

  • The VoltDB server process requires Java 8. The Java client library supports both Java 7 and 8.

  • The required version for the Python client and VoltDB command line u+tilities is 2.6.

18.2.

Improved configuration and reporting for database replication (DR)

As mentioned earlier, the addition of the role attribute to the <dr> element in the configuration file eliminates the need for the SET DR=ACTIVE SQL statement and the --replica argument to the voltdb start command. In addition, two columns for the local and remote cluster IDs have been added to the DRPRODUCER and DRCONSUMER selectors for the @Statistics system procedure. The new columns make it easier to identify to which DR connection the current row of data applies. However, their presence also changes the order of columns from previous releases. If an application used the DRPRODUCER or DRCONSUMER selectors and retrieves column values by index instead of by name, the application will need to be modified to accommodate the new columns in the results.

18.3.

VoltDB now protects against inserting invalid values into TIMESTAMP columns

Previously, VoltDB allowed any 8-byte integer as input to a TIMESTAMP column. However, the TIMESTAMP functions limit the values to those defined by the MIN_VALID_TIMESTAMP() and MAX_VALID_TIMESTAMP() functions. Starting with V7.0, VoltDB no longer accepts values outside of the valid timestamp range for TIMESTAMP columns.

Note that this is an incompatible change and may require changes to applications that are inadvertently inserting invalid timestamp values. If so, you can use the IS_VALID_TIMESTAMP() function to validate input values before inserting them into TIMESTAMP columns.

18.4.

Additional improvements

In addition to the new features and capabilities described above, the following limitations in previous versions have been resolved:

  • There was a race condition where occasionally, when starting XDCR, the request for a synchronizing snapshot from the consumer cluster resulted in an error on the producer cluster indicating it could not initiate the snapshot This issue has been resolved.

  • An issue with elastic expansion of VoltDB clusters (that is, adding nodes to a cluster "on the fly"), was introduced in V6.9. Attempting to add a node could result in the join operation hanging. This did not happen all the time, but has now been corrected and is no longer an issue in V7.0.

  • Previously, the VoltDB export subsystem was not properly releasing all memory from the Java heap during schema or configuration changes. With frequent updates to the schema or configuration, this unused memory could, potentially, accumulate until the Java heap was exhausted. This issue has been resolved.

Known Limitations

The following are known limitations to the current release of VoltDB. Workarounds are suggested where applicable. However, it is important to note that these limitations are considered temporary and are likely to be corrected in future releases of the product.

1. Command Logging

1.1.

Do not use the subfolder name "segments" for the command log snapshot directory.

VoltDB reserves the subfolder "segments" under the command log directory for storing the actual command log files. Do not add, remove, or modify any files in this directory. In particular, do not set the command log snapshot directory to a subfolder "segments" of the command log directory, or else the server will hang on startup.

2. Database Replication

2.1.

Some DR data may not be delivered if master database nodes fail and rejoin in rapid succession.

Because DR data is buffered on the master database and then delivered asynchronously to the replica, there is always the danger that data does not reach the replica if a master node stops. This situation is mitigated in a K-safe environment by all copies of a partition buffering on the master cluster. Then if a sending node goes down, another node on the master database can take over sending logs to the replica. However, if multiple nodes go down and rejoin in rapid succession, it is possible that some buffered DR data — from transactions when one or more nodes were down — could be lost when another node with the last copy of that buffer also goes down.

If this occurs and the replica recognizes that some binary logs are missing, DR stops and must be restarted.

To avoid this situation, especially when cycling through nodes for maintenance purposes, the key is to ensure that all buffered DR data is transmitted before stopping the next node in the cycle. You can do this using the @Statistics system procedure to make sure the last ACKed timestamp (using @Statistitcs DR on the master cluster) is later than the timestamp when the previous node completed its rejoin operation.

2.2.

Avoid bulk data operations within a single transaction when using database replication

Bulk operations, such as large deletes, inserts, or updates are possible within a single stored procedure. However, if the binary logs generated for DR are larger than 45MB, the operation will fail. To avoid this situation, it is best to break up large bulk operations into multiple, smaller transactions. A general rule of thumb is to multiply the size of the table schema by the number of affected rows. For deletes and inserts, this value should be under 45MB to avoid exceeding the DR binary log size limit. For updates, this number should be under 22.5MB (because the binary log contains both the starting and ending row values for updates).

2.3.

Database replication ignores resource limits

There are a number of VoltDB features that help manage the database by constraining memory size and resource utilization. These features are extremely useful in avoiding crashes as a result of unexpected or unconstrained growth. However, these features could interfere with the normal operation of DR when passing data from one cluster to another, especially if the two clusters are different sizes. Therefore, as a general rule of thumb, DR overrides these features in favor of maintaining synchronization between the two clusters.

Specifically, DR ignores any resource monitor limits defined in the deployment file when applying binary logs on the consumer cluster. DR also ignores any partition row limits defined in the database schema when applying binary logs. This means, for example, if the replica database in passive DR has less memory or fewer unique partitions than the master, it is possible that applying binary logs of transactions that succeeded on the master could cause the replica to run out of memory. Note that these resource monitor and tables row limits are applied on any original transactions local to the cluster (for example, transactions on the master database in passive DR).

2.4.

Different cluster sizes can require additional Java heap

Database Replication (DR) now supports replication across clusters of different sizes. However, if the replica cluster is smaller than the master cluster, it may require a significantly larger Java heap setting. Specifically, if the replica has fewer unique partitions than the master, each partition on the replica must manage the incoming binary logs from more partitions on the master, which places additional pressure on the Java heap.

A simple rule of thumb is that the worst case scenario could require an additional P * R * 20MB space in the Java heap , where P is the number of sites per host on the replica server and R is the ratio of unique partitions on the master to partitions on the replica. For example, if the master cluster is 5 nodes with 10 sites per host and a K factor of 1 (i.e. 25 unique partitions) and the replica cluster is 3 nodes with 8 sites per host and a K factor of 1 (12 unique partitions), the Java heap on the replica cluster may require approximately 320MB of additional space in the heap:

Sites-per-host * master/replace ratio * 20MB
8 * 25/12 * 20 = ~ 320MB

An alternative is to reduce the size of the DR buffers on the master cluster by setting the DR_MEM_LIMIT Java property. For example, you can reduce the DR buffer size from the default 10MB to 5MB using the VOLTDB_OPTS environment variable before starting the master cluster.

$ export VOLTDB_OPTS="-DDR_MEM_LIMIT=5"

$ voltdb start

Changing the DR buffer limit on the master from 10MB to 5MB proportionally reduces the additional heap size needed. So in the previous example, the additional heap on the replica is reduced from 320MB to 160MB.

3. Cross Datacenter Replication (XDCR)

3.1.

Avoid replicating tables without a unique index.

Part of the replication process for XDCR is to verify that the record's starting and ending states match on both clusters, otherwise known as conflict resolution. To do that, XDCR must find the record first. Finding uniquely indexed records is efficient; finding non-unique records is not and can impact overall database performance.

To make you aware of possible performance impact, VoltDB issues a warning if you declare a table as a DR table and it does not have a unique index.

3.2.

When starting XDCR for the first time, only one database can contain data.

You cannot start XDCR if both databases already have data in the DR tables. Only one of the two participating databases can have preexisting data when DR starts for the first time.

3.3.

During the initial synchronization of existing data, the receiving database is paused.

When starting XDCR for the first time, where one database already contains data, a snapshot of that data is sent to the other database. While receiving and processing that snapshot, the receiving database is paused. That is, it is in read-only mode. Once the snapshot is completed and the two database are synchronized, the receiving database is automatically unpaused, resuming normal read/write operations.

3.4.

A large number of multi-partition write transactions may interfere with the ability to restart XDCR after a cluster stops and recovers.

Normally, XDCR will automatically restart where it left off after one of the clusters stops and recovers from its command logs (using the voltdb recover command). However, if the workload is predominantly multi-partition write transactions, a failed cluster may not be able to restart XDCR after it recovers. In this case, XDCR must be restarted from scratch, using the content from one of the clusters as the source for synchronizing and recreating the other cluster (using the voltdb create --force command) without any content in the DR tables.

3.5.

A TRUNCATE TABLE transaction will be reported as a conflict with any other write operation to the same table.

When using XDCR, if the binary log from one cluster includes a TRUNCATE TABLE statement and the other cluster performs any write operation to the same table before the binary log is processed, the TRUNCATE TABLE operation will be reported as a conflict. Note that currently DELETE operations always supercede other actions, so the TRUNCATE TABLE will be executed on both clusters.

3.6.

Exceeding a LIMIT PARTITION ROWS constraint can generate multiple conflicts

It is possible to place a limit on the number of rows that any partition can hold for a specific table using the LIMIT PARTITION ROWS clause of the CREATE TABLE statement. When close to the limit, transactions on either or both clusters can exceed the limit simultaneously, resulting in a potentially large number of delete operations that then generate conflicts when the the associated binary log reaches the other cluster.

3.7.

Use of the VoltProcedure.getUniqueId method is unique to a cluster, not across clusters.

VoltDB provides a way to generate a deterministically unique ID within a stored procedure using the getUniqueId method. This method guarantees uniqueness within the current cluster. However, the method could generate the same ID on two distinct database clusters. Consequently, when using XDCR, you should combine the return values of VoltProcedure.getUniqueId with VoltProcedure.getClusterId, which returns the current cluster's unique DR ID, to generate IDs that are unique across all clusters in your environment.

3.8.

XDCR cannot be used with deprecated export syntax.

You cannot use cross datacenter replication (XDCR) with the deprecated export syntax, that is the EXPORT TABLE statement. To use XDCR with export, you must use the current CREATE STREAM syntax for declaring the source streams for export targets.

4. Export

4.1.

Synchronous export in Kafka can use up all available file descriptors and crash the database.

A bug in the Apache Kafka client can result in file descriptors being allocated but not released if the producer.type attribute is set to "sync" (which is the default). The consequence is that the system eventually runs out of file descriptors and the VoltDB server process will crash.

Until this bug is fixed, use of synchronous Kafka export is not recommended. The workaround is to set the Kafka producer.type attribute to "async" using the VoltDB export properties.

5. Import

5.1.

Data may be lost if a Kafka broker stops during import.

If, while Kafka import is enabled, the Kafka broker that VoltDB is connected to stops (for example, if the server crashes or is taken down for maintenance), some messages may be lost between Kafka and VoltDB. To ensure no data is lost, we recommend you disable VoltDB import before taking down the associated Kafka broker. You can then re-enable import after the Kafka broker comes back online.

5.2.

Kafka import can lose data if multiple nodes stop in succession.

There is an issue with the Kafka importer where, if multiple nodes in the cluster fail and restart, the importer can lose track of some of the data that was being processed when the nodes failed. Normally, these pending imports are replayed properly on restart. But if multiple nodes fail, it is possible for some in-flight imports to get lost. This issue will be addressed in an upcoming release.

6. SQL and Stored Procedures

6.1.

Comments containing unmatched single quotes in multi-line statements can produce unexpected results.

When entering a multi-line statement at the sqlcmd prompt, if a line ends in a comment (indicated by two hyphens) and the comment contains an unmatched single quote character, the following lines of input are not interpreted correctly. Specifically, the comment is incorrectly interpreted as continuing until the next single quote character or a closing semi-colon is read. This is most likely to happen when reading in a schema file containing comments. This issue is specific to the sqlcmd utility.

A fix for this condition is planned for an upcoming point release

6.2.

Do not use assertions in VoltDB stored procedures.

VoltDB currently intercepts assertions as part of its handling of stored procedures. Attempts to use assertions in stored procedures for debugging or to find programmatic errors will not work as expected.

6.3.

The UPPER() and LOWER() functions currently convert ASCII characters only.

The UPPER() and LOWER() functions return a string converted to all uppercase or all lowercase letters, respectively. However, for the initial release, these functions only operate on characters in the ASCII character set. Other case-sensitive UTF-8 characters in the string are returned unchanged. Support for all case-sensitive UTF-8 characters will be included in a future release.

7. Client Interfaces

7.1.

Avoid using decimal datatypes with the C++ client interface on 32-bit platforms.

There is a problem with how the math library used to build the C++ client library handles large decimal values on 32-bit operating systems. As a result, the C++ library cannot serialize and pass Decimal datatypes reliably on these systems.

Note that the C++ client interface can send and receive Decimal values properly on 64-bit platforms.

8. SNMP

8.1.

Enabling SNMP traps can slow down database startup.

Enabling SNMP can take up to 2 minutes to complete. This delay does not always occur and can vary in length. If SNMP is enabled when the database server starts, the delay occurs after the server logs the message "Initializing SNMP" and before it attempts to connect to the cluster. If you enable SNMP while the database is running, the delay can occur when you issue the voltadmin update command or modify the setting in the VoltDB Management Center Admin tab. This issue results from a Java constraint related to secure random numbers used by the SNMP library.

9. VoltDB Management Center

9.1.

The VoltDB Management Center currently reports on only one DR connection.

With VoltDB V7.0, cross datacenter replication (XDCR) supports multiple clusters in an XDCR network. However, the VoltDB Management Center currently reports on only one such connection per cluster. In the future, the Management Center will provide monitoring and statistics for all connections to the current cluster.

Implementation Notes

The following notes provide details concerning how certain VoltDB features operate. The behavior is not considered incorrect. However, this information can be important when using specific components of the VoltDB product.

1. VoltDB Management Center

1.1.

Schema updates clear the stored procedure data table in the Management Center Monitor section

Any time the database schema or stored procedures are changed, the data table showing stored procedure statistics at the bottom of the Monitor section of the VoltDB Management Center get reset. As soon as new invocations of the stored procedures occur, the statistics table will show new values based on performance after the schema update. Until invocations occur, the procedure table is blank.

2. SQL

2.1.

You cannot partition a table on a column defined as ASSUMEUNIQUE.

The ASSUMEUNIQUE attribute is designed for identifying columns in partitioned tables where the column values are known to be unique but the table is not partitioned on that column, so VoltDB cannot verify complete uniqueness across the database. Using interactive DDL, you can create a table with a column marked as ASSUMEUNIQUE, but if you try to partition the table on the ASSUMEUNIQUE column, you receive an error. The solution is to drop and add the column using the UNIQUE attribute instead of ASSUMEUNIQUE.

2.2.

Adding or dropping column constraints (UNIQUE or ASSUMEUNIQUE) is not supported by the ALTER TABLE ALTER COLUMN statement.

You cannot add or remove a column constraint such as UNIQUE or ASSUMEUNIQUE using the ALTER TABLE ALTER COLUMN statement. Instead to add or remove such constraints, you must first drop then add the modified column. For example:

ALTER TABLE employee DROP COLUMN empID;
ALTER TABLE employee ADD COLUMN empID INTEGER UNIQUE;

2.3.

Do not use UPDATE to change the value of a partitioning column

For partitioned tables, the value of the column used to partition the table determines what partition the row belongs to. If you use UPDATE to change this value and the new value belongs in a different partition, the UPDATE request will fail and the stored procedure will be rolled back.

Updating the partition column value may or may not cause the record to be repartitioned (depending on the old and new values). However, since you cannot determine if the update will succeed or fail, you should not use UPDATE to change the value of partitioning columns.

The workaround, if you must change the value of the partitioning column, is to use both a DELETE and an INSERT statement to explicitly remove and then re-insert the desired rows.

2.4.

Certain SQL syntax errors result in the error message "user lacks privilege or object not found".

If you refer to a table or column name that does not exist, VoltDB reports that the "user lacks privilege or object not found". This can happen, for example, if you misspell a table or column name.

Another situation where this occurs is if you mistakenly use double quotation marks to enclose a string literal (such as WHERE ColumnA="True"). ANSI SQL requires single quotes for string literals and reserves double quotes for object names. In the preceding example, VoltDB interprets "True" as an object name, cannot resolve it, and issues the "user lacks privilege" error.

The workaround is, if you receive this error, to look for misspelled table or columns names or string literals delimited by double quotes in the offending SQL statement.

2.5.

Ambiguous column references no longer allowed.

Starting with VoltDB 6.0, ambiguous column references are no longer allowed. For example, if both the Customer and Placedorder tables have a column named Address, the reference to Address in the following SELECT statement is ambiguous:

SELECT OrderNumber, Address FROM Customer, Placedorder
   . . .

Previously, VoltDB would select the column from the leftmost table (Customer, in this case). Ambiguous column references are no longer allowed and you must use table prefixes to disambiguate identical column names. For example, specifying the column in the preceding statement as Customer.Address.

A corollary to this change is that a column declared in a USING clause can now be referenced using a prefix. For example, the following statement uses the prefix Customer.Address to disambiguate the column selection from a possibly similarly named column belonging to the Supplier table:

SELECT OrderNumber, Vendor, Customer.Address
   FROM Customer, Placedorder Using (Address), Supplier
    . . .
3. Runtime

3.1.

File Descriptor Limits

VoltDB opens a file descriptor for every client connection to the database. In normal operation, this use of file descriptors is transparent to the user. However, if there are an inordinate number of concurrent client connections, or clients open and close many connections in rapid succession, it is possible for VoltDB to exceed the process limit on file descriptors. When this happens, new connections may be rejected or other disk-based activities (such as snapshotting) may be disrupted.

In environments where there are likely to be an extremely large number of connections, you should consider increasing the operating system's per-process limit on file descriptors.

3.2.

Use of Resources in JAR Files

There are two ways to access additional resources in a VoltDB database. You can place the resources in the /lib folder where VoltDB is installed on each server in the cluster or you can include the resource in a subfolder of a JAR file you add using the sqlcmd LOAD CLASSES directive. Adding resources via the /lib directory is useful for stable resources (such as third-party software libraries) that do not require updating. Including resources (such as XML files) in the JAR file is useful for resources that may need to be updated, as a single transaction, while the database is running.

LOAD CLASSES is used primarily to load classes associated with stored procedures and user-defined functions. However, it will also load any additional resource files included in subfolders of the JAR file. You can remove classes that are no longer needed using the REMOVE CLASSES directive. However, there is no explicit command for removing other resources.

Consequently, if you rename resources or move them to a different location and reload the JAR file, the database will end up having multiple copies. Over time, this could result in more and more unnecessary memory being used by the database. To remove obsolete resources, you must first reinitialize the database root directory, start a fresh database, reload the schema (including the new JAR files with only the needed resources) and then restore the data from a snapshot.