Percona XtraDB Cluster moved to https://jira.percona.com/projects/PXC

  1. Bug #1260713
  2. Comment #5

Comment 5 for bug 1260713

Revision history for this message
Frederic Descamps (lefred) wrote :

Just an extra comment:

It's very clear from the slow logs and binary logs that on the master a transaction deletes many rows in multiple tables (up to 1k rows or more) and after that a second transaction deletes only one row.

I suspect that on the "slave" as it fails only when wsrep_slave_threads > 1, the second transaction is executed and committed faster.

I don't know if the writeset issue with FK is related of the amount of rows or tables involved.