Apache Accumulo 2.1.1
02 Nov 2022
These are draft release notes for a future release of Accumulo!
Please view the latest release notes.
This version is a current Long Term Maintenance (LTM) release.
Apache Accumulo 2.1.1 is a patch release of the 2.1.0 LTM line.
- #3180 Enable users to provide per-volume Hadoop Filesystem configuration overrides via the Accumulo configuration. Hadoop Filesystem objects are configured by the standard Hadoop mechanisms (default configuration, core-site.xml, hdfs-site.xml, etc.), but these configuration files don’t allow for the same property to be specified with different values for different namespaces. This change allows users to specify different property values for different Accumulo volumes, which will be applied to the Hadoop Filesystem object created for each Accumulo volume.
- #3177 Added server side code to validate that property values conform to the specified property type (string, boolean, etc.).
- #3175 Reset number of locks in SynchronousLoadingBlockCache from 2017 back to 5003, the value that it was in 1.10. Also, modified the lock to be fair, which allows the different scan threads in the server to make progress in a more fair manner when they need to load a block into the cache.
- #3118 Added option to the
admin zoo-info-viewercommand to dump the ACLs on ZooKeeper nodes. This information can be used to fix znodes with incorrect ACLs during the upgrade process.
- #3077, #3079, #3083, #3123 Avoid filling
OS page cache by calling
setDropBehindon the FS data stream when performing likely one-time file accesses, as with WAL and compaction input and output files. This should allow files that might benefit more from caching to stay in the cache longer. #3083 and #3123 introduces new properties, table.compaction.major.output.drop.cache and table.compaction.minor.output.drop.cache, for dropping pages from the OS page cache for compaction output files. These changes will only have an impact on HDFS FileSystem implementations and operating systems that support the underlying OS system call. See associated HDFS issue that will improve the underlying implementation when resolved.
- #3023, #3053 Prevent excessive logging when an ExternalCompaction is cancelled for cause and around the ThreadPools
- #3057 Modified Monitor to direct DataTable errors to console
- #3205 Support writing cluster config parser output to a file
- #3164 Fixed bug in ScanServer where failures to scan a specific tablet was not being handled correctly, which would lead to the tablet not being scanned.
- #3059 Removed stale compactions from CompactionCoordinator internal structure. The CompactionCoordinator has an internal data structure where it keeps track of running compactions. Running compactions were not being removed from this set in some edge cases so they continued to be shown on the Monitor when in reality the compaction was not running. This occured in testing when the Compactor process was killed by the agitator. The compaction was rescheduled to run on a different Compactor, the issue here was just orphaned information in the data structure being shown in the Monitor.
- #3048 Output time strings on the monitor using the correct user locale (also made them more concise)
View the Upgrading Accumulo documentation for guidance.
2.1.1 GitHub Project
At the time of release, the following issues were known:
View all releases in the archive