Actions
Bug #38598
closedosdmap may include only v1 address while osd binds to v2; mon drops messages
% Done:
0%
Source:
Tags:
Backport:
Regression:
No
Severity:
3 - minor
Reviewed:
Affected Versions:
ceph-qa-suite:
Component(RADOS):
Pull request ID:
Crash signature (v1):
Crash signature (v2):
Description
- osd binds to v2+v1
- osd sends osd_boot to mon
- mon adds v1 addr to osdmap only (due to require_osd_release < nautilus, perhaps)
- osd sends random messages to mon, like MOSDAlive, and mon drops them because the get_source_addrs() (v2+v1) doesn't match the map
This appears to be causing the upgrade/mimic-x and luminous-x failures.
Updated by Sage Weil about 5 years ago
Proposed OSD fix:
- if we get an osdmap with require_osd_release < nautilus, and are bound to v2+v1, we rebind to v1 only.
- if the cluster later allows v2 (due to change in require_osd_release), too bad. restart your osd.
Updated by Neha Ojha about 5 years ago
- Status changed from 12 to Fix Under Review
- Pull request ID set to 26794
Updated by Sage Weil about 5 years ago
- Status changed from Fix Under Review to Resolved
Actions