OK, I'll create an issue for this.
Though my comment is that both of the systems involved (docker registry
and Hawkular metrics) are core parts of openshift so I would hole that
no "fiddling" would be needed.
On 05/07/18 15:57, Ben Parees wrote:
I forwarded your problem on to our storage team lead, he had the
following suggestions:
"I believe they will want to fiddle with the fsGroup or
supplementalGroup so that it matches the GID of the cassandra user and
make sure those GIDs are in the SCC ranges for the pod."
He also recommended you consider opening a bugzilla as it's easier to
track these issues that way.
On Thu, Jul 5, 2018 at 7:42 AM, Tim Dudgeon <tdudgeon...@gmail.com
<mailto:tdudgeon...@gmail.com>> wrote:
I hit this problem again, this time with the cassandra pod for
Hawkular metrics.
This has been running without problem for some months, but now I
found that the cassandra pod could not start because of file
permissions writing to the /cassandra_data/data directory.
Looking at that directory the ownership was set to
1000040000.65534, but cassandra was running as user 313 so could
not write to that directory. Manually changing permissions to
313.65534 (the 65534 group is nfsnobody, and the cassandra user is
a member of that group) fixed the problem and allowed the
cassandra pod to start.
Clearly the 1000040000 user is an openshift assigned user, but as
the container is running as the cassandra user (313) I have no
idea how this could have happened.
Can anyone explain what is going on here?
Tim
On 02/07/18 16:27, Tim Dudgeon wrote:
I've hit a strange problem with directory ownership for the
docker registry a couple of times, and don't understand what
is causing this.
The registry was working fine for some time. I'm using a
Cinder volume for the registry storage, but don't know if
that's relevant.
Then something happened that stopped pods pushing to the
registry, with the problem being that the registry pod was
getting "Permission denied" errors when it was trying to
create directories under
/registry/docker/registry/v2/repositories.
Looking at the file system the directories were all owned by
1000000000.1000000000 which explains why the registry process
(running as user 1001) could not write to these directories. e.g.
sh-4.2$ cd /registry/docker/registry/v2/
sh-4.2$ ls -al
total 0
drwxrwsr-x. 4 1000000000 1000000000 39 Apr 20 15:51 .
drwxrwsr-x. 3 1000000000 1000000000 16 Apr 20 15:51 ..
drwxrwsr-x. 3 1000000000 1000000000 20 Apr 20 15:51 blobs
drwxrwsr-x. 15 1000000000 1000000000 215 May 29 14:14 repositories
Doing a `docker -exec -u 0 <registry-pod> on the infra node
and then a `chown -R 1001.0 /registry/docker/registry` to
reset the permissions fixed the problem.
Anyone any idea what's going on here?
Tim
_______________________________________________
users mailing list
users@lists.openshift.redhat.com
<mailto:users@lists.openshift.redhat.com>
http://lists.openshift.redhat.com/openshiftmm/listinfo/users
<http://lists.openshift.redhat.com/openshiftmm/listinfo/users>
--
Ben Parees | OpenShift
_______________________________________________
users mailing list
users@lists.openshift.redhat.com
http://lists.openshift.redhat.com/openshiftmm/listinfo/users