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

  1. Bug #1277703
  2. Comment #0

Comment 0 for bug 1277703

Revision history for this message
Jay Janssen (jay-janssen) wrote : 5.6 cluster node that is a slave with parallel worker threads seems to generate writesets that are not parallelizable

I don't have this fully locked down, but I suspect there may be something here:

Scenario: 2 node cluster, first node has 10 --slave-parallel-threads in inbound async replication. No other writes, wsrep_preordered=OFF.

2nd node starts and gets stuck in JOINED catching up very slowly, 80-100 tps.
2nd node has a persistent wsrep_cert_deps_distance=1.0000

1st node is very fast by itself: 1k+ tps, but 2nd node JOINED sends a lot of FC.

My diagnosis is that the send node can't keep up because the writesets are not allowing parallelism for some reason. My hunch is that this may be related to slave-parallel-workers.