Actions
Bug #12688
closedpg import can cause bogus interval (first > last)
Status:
Duplicate
Priority:
Urgent
Assignee:
-
Category:
-
Target version:
-
% Done:
0%
Source:
Q/A
Tags:
Backport:
Regression:
No
Severity:
3 - minor
Reviewed:
Affected Versions:
ceph-qa-suite:
Pull request ID:
Crash signature (v1):
Crash signature (v2):
Description
remote/plana19/log/ceph-osd.4.log.gz:2015-08-12 20:30:10.428586 7f103f848700 10 osd.4 pg_epoch: 745 pg[0.c1( empty local-les=0 n=0 ec=1 les/c 726/726 737/759/737) [3] r=-1 lpr=745 pi=726-744/8 crt=0'0 inactive] start_peering_interval: check_new_interval output: generate_past_intervals interval(759-744 up [4,3](4) a cting [4,3](4)): not rw, up_thru 737 up_from 657 last_epoch_clean 726 remote/plana19/log/ceph-osd.4.log.gz:2015-08-12 20:30:10.428598 7f103f848700 10 osd.4 pg_epoch: 745 pg[0.c1( empty local-les=0 n=0 ec=1 les/c 726/726 737/759/737) [3] r=-1 lpr=745 pi=726-744/8 crt=0'0 inactive] noting past interval(759-744 up [4,3](4) acting [4,3](4)) remote/plana19/log/ceph-osd.4.log.gz:2015-08-12 20:30:10.428609 7f103f848700 10 osd.4 pg_epoch: 745 pg[0.c1( empty local-les=0 n=0 ec=1 les/c 726/726 745/745/745) [3] r=-1 lpr=745 pi=726-744/8 crt=0'0 inactive] up [4,3] -> [3], acting [4,3] -> [3], acting_primary 4 -> 3, up_primary 4 -> 3, role 0 -> -1
the pg was just exported from another OSD with a newer OSDMap. the new osd gets confused by the intervals and info.history, which reference future epochs it hasn't seen yet.
/a/sage-2015-08-12_14:04:07-rados-wip-sage-testing---basic-multi/1012335
Actions