Setup:
3 node percona pxc cluster with the following percona images
Operator: 1.19.0
Cluster: 8.0.42-33.1
Xtrabackup: 8.0.35-34.1
I was trying PITR, during which the restore job succeeded, but the pitr job failed because it hit the max_allowed_packet limit.
I had a couple of questions:
- The PITR container didn’t exit after failing and I had to manually delete the pod for it to retry, is there a way to avoid this?
- How can i prevent hitting the max_packet_size limit while performing PITR since the max value for max_allowed_packet is 1Gb and we have max txn size as 2Gb?
@Yash_Daga for 1, do you have pitr job logs?
Created Jira to track this.
Unfortunately, Atlassian doesn’t provide the necessary data via the link to unauthorized users (or websites/bots), and there’s no plugin for this in Discourse, so I set it up so that at least the full URL is displayed.
https://perconadev.atlassian.net/browse/K8SPXC-1876
Hi @Ege_Gunes thanks for the response.
Unfortunately, I don’t have those logs currently, I’ll try to see if I can retry this workflow and get this error.
I did found this from back when we got this error, though:
mysql> SELECT logged, prio, error_code, data FROM performance_schema.error_log ORDER BY logged DESC LIMIT 20\G *************************** 1. row *************************** logged: 2026-04-17 13:00:51.213911 prio: Note error_code: MY-010914 data: Aborted connection 2486 to db: 'unconnected' user: 'xtrabackup' host: 'mysqlcluster-pxc-0.mysqlcluster-pxc.percona-operator.svc.cluster.local' (Got a packet bigger than 'max_allowed_packet' bytes). *************************** 2. row *************************** logged: 2026-04-17 13:00:42.417551 prio: Note error_code: MY-010914 data: Aborted connection 69 to db: 'mysql' user: 'xtrabackup' host: 'mysqlcluster-pxc-0.mysqlcluster-pxc.percona-operator.svc.cluster.local' (Got an error reading communication packets).
pitr job hanging is worrying, I’ll try to simulate going over max packet size to see what happens.