Actions
Bug #38606
openceph-mgr tries to read a ceph.conf after a respawn
Status:
New
Priority:
Normal
Assignee:
-
Category:
-
Target version:
-
% Done:
0%
Source:
Tags:
Backport:
Regression:
No
Severity:
3 - minor
Reviewed:
Affected Versions:
ceph-qa-suite:
Pull request ID:
Crash signature (v1):
Crash signature (v2):
Description
In Rook, we don't use a ceph.conf at all, when applying configs to modules (dashboard port/server address for instance) the mgr respawns itself, but fails because it cannot read a ceph.conf:
[leseb@tarox~/go/src/github.com/rook/rook][use-ceph-user !] docker logs ed0b5a159aac debug 2019-03-06 15:40:02.396 7fe79161a380 0 set uid:gid to 167:167 (ceph:ceph) debug 2019-03-06 15:40:02.396 7fe79161a380 0 ceph version 13.2.4 (b10be4d44915a4d78a8e06aa31919e74927b142e) mimic (stable), process ceph-mgr, pid 1 debug 2019-03-06 15:40:02.396 7fe79161a380 0 pidfile_write: ignore empty --pid-file debug 2019-03-06 15:40:02.423 7fe79161a380 1 mgr[py] Loading python module 'balancer' debug 2019-03-06 15:40:02.443 7fe79161a380 1 mgr[py] Loading python module 'dashboard' debug 2019-03-06 15:40:02.502 7fe79161a380 1 mgr[py] Loading python module 'hello' debug 2019-03-06 15:40:02.512 7fe79161a380 1 mgr[py] Loading python module 'influx' debug 2019-03-06 15:40:02.522 7fe79161a380 1 mgr[py] Loading python module 'iostat' debug 2019-03-06 15:40:02.532 7fe79161a380 1 mgr[py] Loading python module 'localpool' debug 2019-03-06 15:40:02.543 7fe79161a380 1 mgr[py] Loading python module 'prometheus' debug 2019-03-06 15:40:02.591 7fe79161a380 1 mgr[py] Loading python module 'restful' debug 2019-03-06 15:40:02.720 7fe79161a380 1 mgr[py] Loading python module 'selftest' debug 2019-03-06 15:40:02.731 7fe79161a380 1 mgr[py] Loading python module 'smart' debug 2019-03-06 15:40:02.741 7fe79161a380 1 mgr[py] Loading python module 'status' debug 2019-03-06 15:40:02.761 7fe79161a380 1 mgr[py] Loading python module 'telegraf' debug 2019-03-06 15:40:02.778 7fe79161a380 1 mgr[py] Loading python module 'telemetry' debug 2019-03-06 15:40:02.833 7fe79161a380 1 mgr[py] Loading python module 'zabbix' debug 2019-03-06 15:40:02.847 7fe79161a380 1 mgr send_beacon standby debug 2019-03-06 15:40:02.848 7fe77d1ea700 1 mgr respawn e: 'ceph-mgr' debug 2019-03-06 15:40:02.848 7fe77d1ea700 1 mgr respawn 0: 'ceph-mgr' debug 2019-03-06 15:40:02.848 7fe77d1ea700 1 mgr respawn 1: '--fsid=9dd7c26c-a7d4-414a-abae-2f6980b18faa' debug 2019-03-06 15:40:02.848 7fe77d1ea700 1 mgr respawn 2: '--keyring=/etc/ceph/keyring-store/keyring' debug 2019-03-06 15:40:02.848 7fe77d1ea700 1 mgr respawn 3: '--log-to-stderr=true' debug 2019-03-06 15:40:02.848 7fe77d1ea700 1 mgr respawn 4: '--err-to-stderr=true' debug 2019-03-06 15:40:02.848 7fe77d1ea700 1 mgr respawn 5: '--mon-cluster-log-to-stderr=true' debug 2019-03-06 15:40:02.848 7fe77d1ea700 1 mgr respawn 6: '--log-stderr-prefix=debug ' debug 2019-03-06 15:40:02.848 7fe77d1ea700 1 mgr respawn 7: '--mon-host=10.98.48.32:6789' debug 2019-03-06 15:40:02.848 7fe77d1ea700 1 mgr respawn 8: '--mon-initial-members=a' debug 2019-03-06 15:40:02.848 7fe77d1ea700 1 mgr respawn 9: '--name=mgr.a' debug 2019-03-06 15:40:02.848 7fe77d1ea700 1 mgr respawn 10: '--setuser=ceph' debug 2019-03-06 15:40:02.848 7fe77d1ea700 1 mgr respawn 11: '--setgroup=ceph' debug 2019-03-06 15:40:02.848 7fe77d1ea700 1 mgr respawn 12: '--foreground' debug 2019-03-06 15:40:02.848 7fe77d1ea700 1 mgr respawn respawning with exe /usr/bin/ceph-mgr debug 2019-03-06 15:40:02.848 7fe77d1ea700 1 mgr respawn exe_path /proc/self/exe global_init: error reading config file.
So basically the config was correctly applied, the container died, and respawned successfully.
We need to not try to read the ceph.conf if the CLI has all the info we need.
For instance Rook runs the mgr like this:
ceph-mgr --fsid=2bfe4553-0a30-4d69-95e6-8a5ab1602080 --keyring=/etc/ceph/keyring-store/keyring --log-to-stderr=true --err-to-stderr=true --mon-cluster-log-to-stderr=true --log-stderr-prefix=debug --mon-host=10.108.10.124:6789 --mon-initial-members=a --name=mgr.a --setuser=ceph --setgroup=ceph --foreground
Updated by Sebastian Wagner almost 5 years ago
IS this specific for MGRs or are all daemons affected?
Updated by Amit Ghadge over 4 years ago
Hi,
I'm facing similar kind of issue with prometheus plugin in my ceph cluster. There is not return more details in "global_init: error reading config file." error.
How to debug in my ceph cluster.
Actions