Resource unavailable after crash
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
MySQL patches by Codership |
Fix Released
|
Medium
|
Alex Yurchenko | ||
codership-maria |
Fix Committed
|
Medium
|
Alex Yurchenko |
Bug Description
After crash which described in bug #797396, when I am trying to start JOINER again, it fails:
110614 13:37:28 [Note] WSREP: Flow-control interval: [12, 23]
110614 13:37:28 [Note] WSREP: Shifting OPEN -> PRIMARY (TO: 8413768)
110614 13:37:28 [Note] WSREP: New cluster view: group UUID: 0b743350-
110614 13:37:28 [Warning] WSREP: Gap in state sequence. Need state transfer.
110614 13:37:30 [Note] [DEBUG] WSREP: Prepared SST request: rsync|10.
110614 13:37:30 [Warning] WSREP: wsrep_notify_cmd is not defined, skipping notification.
110614 13:37:30 [Note] WSREP: State transfer required:
Group state: 0b743350-
Local state: 00000000-
110614 13:37:30 [Note] WSREP: Requesting state transfer failed: -11(Resource temporarily unavailable). Will keep retrying every 1 second(s)
It seems DONOR still keeps open connection.
This is what I see on DONOR:
tcp 0 0 10.11.12.234:4567 10.11.12.220:46621 ESTABLISHED 12187/mysqld
( where 10.11.12.234 is donor and 10.11.12.220 is JOINER).
And this connections stays open after JOINER is crashed or killed.
I do not know how to close this connection without restarting DONOR, which is quite annoyance.
In additional, when JOINER gets into state
"WSREP: Requesting state transfer failed: -11(Resource temporarily unavailable). Will keep retrying every 1 second(s)"
there is no way to shutdown mysqld process, only to kill it - mysqld does not accept connections on this stage.
Changed in codership-mysql: | |
assignee: | nobody → Alex Yurchenko (ayurchen) |
importance: | Undecided → Medium |
milestone: | none → 0.8.1 |
status: | New → Confirmed |
Changed in codership-maria: | |
assignee: | nobody → Alex Yurchenko (ayurchen) |
importance: | Undecided → Medium |
status: | New → Confirmed |
Changed in codership-mysql: | |
status: | Fix Committed → Fix Released |
This should be fixed as part of lp:797396 fix