I am running a CentOS 7 three node XtraDB cluster ususally with yum-cron enabled and usually with no issues.
The repo now contains 5.7.17-27.20.2.el7.x86_64 packages. After yum-cron installed them no two of the nodes, nagios showed “ERROR 1682 (HY000) at line 1: Native table ‘performance_schema’.‘global_variables’ has the wrong structure” on these two.
I disabled yum-cron, downgraded the packages on the nodes in question, rebooted these nodes and things were OK again without the above error. After that, I also ran mysql_upgrade to be on the safe side.
Next step was that I manually upgraded the nodes and ran mysql_upgrade thereafter. That did not make a difference, unfortunately. The error did remain on the same two nodes.
Then, I downgraded the packages again. Unfortunately, the error does now stay. The cluster does seem to run, but nagios output is no longer really useful. Only the single node which was not touched does show cluster size and flow controll (both OK), while the others just show the error in place of these values.
Can someone please shed some light on how to deal with this? Release notes do not seem to be on the website, yet.