Skip to content

NebulaGraph 3.1.3 release notes

Changes

  • Vertices cannot be inserted without a tag by default. To insert a vertex without a tag, add the --graph_use_vertex_key=true option to the nebula-graphd.conf configuration files of all Graph services, and add --use_vertex_key=true to the nebula-storaged.conf configuration files of all Storage services in the cluster.

Enhancement

  • Enhanced memory usage of AtomicLogBuffer to avoid OOM problems when rebuilding indexes and data synchronization.
  • Unified the configuration file.
  • Adjusted the level of the heartbeat log.

Bugfix

  • Fixed the web service crash when receiving some special attack messages.
  • Fixed the crash due to the full-text index dropping.
  • Fixed the crash due to map concurrency.
  • Fixed the raft crash in certain cases.
  • Fixed the storage service crash when the length of the VID in a vertex or edge deletion statement exceeds the defined length on an indexed base.
  • Fixed the crash caused by a wrong expression syntax.
  • Fixed the crash caused by LOOKUP statements with AND and IN operators.
  • Fixed the crash when running a complex MATCH statement.
  • Fixed the crash when running YIELD 1 AS number MINUS YIELD 2 AS number.
  • Fixed the crash in the optimization phase of multiple MATCH.
  • Fixed the crash when collecting variable types.
  • Fixed the crash caused by an illegal expression.
  • Fixed the bug that storage would hang when there is only the space path but no part path.
  • Fixed the deadlock of leader balance job executor.
  • Fixed the infinite loop when building a balance plan.
  • Fixed the failure of rebuilding a full-text index.
  • Fixed the logrotate usage problem.
  • Fixed the bug that the machine key is lost when restoring.
  • Fixed the bug that the host can't stop when canceling the progress of sending snapshots.
  • Fixed the cache size overflow and deadlock bug.
  • Fixed the raft split brain problem.
  • Fixed the bug that the meta listener validated the license.
  • Fixed the bug that the meta listener didn't clean data.
  • Fixed the bug that the drainer synchronized dirty data.
  • Fixed the bug that the drainer daemon can't exit normally.
  • Fixed the problem that audit log cannot be used asynchronously.
  • Fixed the concurrent exception when executing multiple MATCH statements.
  • Fixed the bug that the tag index rebuilding task could not be re-executed properly.
  • Fixed the bug that the tag index rebuilding task always failed after stopping a running job.
  • Fixed the bug that the ElasticSearch writing error caused by truncation of UTF8 characters.
  • Fixed the bug that the truncated text was removed before writing into ElasticSearch.
  • Fixed the bug that the audit logs of DML and DQL types are not recorded when using ElasticSearch to store audit logs.
  • Fixed the bug that the service could not be started if the log directory did not exist when ENABLE_BREAKPAD was enabled.
  • Fixed the bug that when there was a GOD whose name was not root, the root user were created when the meta initiated.
  • Fixed the backup failure after dropping a space.
  • Fixed the TTL non-handling problem when scanning vertices or edges.
  • Fixed the time inconsistency when running INSERT and checking the valid of the TTL field.
  • Fixed the bug that the synchronization between two clusters may fail when the vid type was string.

Legacy versions

Release notes of legacy versions

Last update: February 19, 2024