Bug #20219: upgrade/jewel-x: ceph_test_rados_api_stat hangs on getpoolstats - Ceph - Ceph

Actions

Copy link

Bug #20219

closed

upgrade/jewel-x: ceph_test_rados_api_stat hangs on getpoolstats

Added by Sage Weil almost 7 years ago. Updated almost 7 years ago.

Status:

Resolved

Priority:

Immediate

Assignee:

Category:

Target version:

% Done:

Source:

Tags:

Backport:

Regression:

Severity:

3 - minor

Reviewed:

Affected Versions:

ceph-qa-suite:

Pull request ID:

Crash signature (v1):

Crash signature (v2):

Description

2017-06-07 20:48:21.229431 7f34d82bb700  1 -- 172.21.15.18:6789/0 <== client.64639 172.21.15.18:0/2471129480 10 ==== getpoolstats(15 test-rados-api-smithi018-4030-1 v1677) v1 ==== 73+0+0 (2354001487 0 0) 0x7f34ef573a80 con 0x7f34eff14800
2017-06-07 20:48:21.229447 7f34d82bb700 20 mon.b@1(peon) e2 _ms_dispatch existing session 0x7f34f0086700 for client.? 172.21.15.18:0/2471129480
2017-06-07 20:48:21.229451 7f34d82bb700 20 mon.b@1(peon) e2  caps allow *
2017-06-07 20:48:21.229453 7f34d82bb700 10 mon.b@1(peon).paxosservice(mgrstat 1..1) dispatch 0x7f34ef573a80 getpoolstats(15 test-rados-api-smithi018-4030-1 v1677) v1 from client.64639 172.21.15.18:0/2471129480 con 0x7f34eff14800
2017-06-07 20:48:21.229458 7f34d82bb700 10 mon.b@1(peon).paxosservice(mgrstat 1..1)  waiting for paxos -> readable (v1677)

but paxos is readable. However, the mgrstat svc is at 1 and getpoolstats has v1677 from the old pgmon.

/a/sage-2017-06-07_20:02:25-upgrade:jewel-x-wip-sage-testing2---basic-smithi/1268343

Actions

Copy link

Updated by Sage Weil almost 7 years ago

a few ways to fix this:

1. make mgrstatmonitor ignore the version entirely
2. somehow make mgrstatmonitor advance to the pgmon version when the switchover happens.
3. make mgrstatmonitor take over the paxos slot and prefix when the swithover happens (e.g., change pgmon() to return MgrStatMonitor instead of PGMonitor)).

whatever we do, we have to make sure that a running client, that is remembering the paxos v from pgmon on jewel, will continue to work without restarting when the mons upgrade to luminous.

Actions

Copy link