Conflicting Transaction Sets following Complete Outage of InnoDB Cluster

Bug #1854880 reported by David Ames
14
This bug affects 2 people
Affects Status Importance Assigned to Milestone
MySQL Server
Unknown
Unknown
mysql-8.0 (Ubuntu)
Fix Released
Critical
Unassigned

Bug Description

[Impact]
The 8.0.18 version of mysql-8.0 has this upstream bug:

https://bugs.mysql.com/bug.php?id=97279

Note 8.0.17 did not display this bug.

After a complete outage dba.rebootClusterFromCompleteOutage() errors with:

output: "ERROR: Conflicting transaction sets between 10.5.0.55:3306 and 10.5.0.52:3306 Dba.rebootClusterFromCompleteOutage: Conflicting transaction sets between 10.5.0.55:3306 and 10.5.0.52:3306 (MYSQLSH 51152) at /tmp/tmpn2y5qo9l.js:3:12 in dba.rebootClusterFromCompleteOutage();

[Regression Potential]
TBD

[Fix]

[Test Case]
* Build an InnoDB cluster
  - var cluster = dba.createCluster('cluster')
  - cluster.addInstance('<email address hidden>:xxxx')
  - cluster.addInstance('<email address hidden>:yyyy')
  - cluster.setOption("consistency", "BEFORE")
* Shut each node of the cluster down (reboot)
* Start mysql on each node
* (Maybe?) INSERT some random data into the same table on all cluster instances
* Run dba.rebootClusterFromCompleteOutage()

The cs:~openstack-charmers-next/mysql-innodb-cluster has a test zaza.openstack.charm_tests.mysql.tests.MySQLInnoDBClusterColdStartTest that re-creates the failure. Note: It is currently disabled due to this bug.

[Workaround]
From upstream bug report,
"""
  a) Shutdown the instance that contains the GTID subset (typically a secondary that was behind, or a primary that was ejected from the cluster)
  b) Execute rebootClusterFromCompleteOutage to bring the cluster up
  c) Start the instance from step a)
  d) Execute a cluster.rescan() to bring the instance back into the cluster
"""

Robie Basak (racb)
Changed in mysql-8.0 (Ubuntu):
status: New → Triaged
importance: Undecided → Critical
tags: added: regression-update
Revision history for this message
Bryce Harrington (bryce) wrote :

A patch is not yet available for this issue from upstream, however it is confirmed by them with their own test case defined, and a workaround, both of which I've updated in the bug description. I've linked the upstream bug to hopefully give visibility when a patch does become available.

description: updated
Revision history for this message
Alex Lutay (taurus) wrote (last edit ):

The issue can be resolved, as upstream fixed it in 8.0.19 and the current MySQL version is 8.0.33 in Ubuntu[security].

Lena Voytek (lvoytek)
Changed in mysql-8.0 (Ubuntu):
status: Triaged → Fix Released
To post a comment you must log in.
This report contains Public information  
Everyone can see this information.

Other bug subscribers

Remote bug watches

Bug watches keep track of this bug in other bug trackers.