Cluster nodes sensitive to time differences

Bug #832305 reported by Shawn Usry
6
This bug affects 1 person
Affects Status Importance Assigned to Milestone
MySQL patches by Codership
Triaged
Undecided
Unassigned
0.8
Triaged
Undecided
Unassigned

Bug Description

I'm not sure if this is a bug, or maybe just an observation that just needs further highlight in requirements / documentation (apologies if it's there and I missed it!).

I noticed in my 2-node setup, that my state transfers would succeed, but my "Joiner" node would not move to an operational state (on) if the time set on each node was out of difference by more than a few minutes. Mine were off by about 5-6 minutes Restarts of the joiner node would not resolve this. I didn't do extensive testing on the exact deltas, but as soon as I properly updated my time offsets from a time server, the very next state transfer succeeded and brought the node to useable state.

Revision history for this message
Teemu Ollakka (teemu-ollakka) wrote :

Could not reproduce this, configured two servers with 30 min time difference and verified that nodes can be joined with state transfer in both directions.

To investigate this further, more information is required:
* What SST method was in use (mysqldump or rsync)?
* Logs from both joiner and donor, starting from "New cluster view..." notification

If this can be reproduced easily, it would help to have output of the following commands on both joiner and donor:
* show processlist
* show status like 'wsrep%'

Changed in codership-mysql:
status: New → Triaged
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.