Bug #12778
closedepel repo issues
0%
Description
2015-08-20T21:58:05.752 INFO:teuthology.orchestra.run.plana50.stderr: One of the configured repositories failed (Unknown), 2015-08-20T21:58:05.752 INFO:teuthology.orchestra.run.plana50.stderr: and yum doesn't have enough cached data to continue. At this point the only 2015-08-20T21:58:05.752 INFO:teuthology.orchestra.run.plana50.stderr: safe thing yum can do is fail. There are a few ways to work "fix" this: 2015-08-20T21:58:05.752 INFO:teuthology.orchestra.run.plana50.stderr: 2015-08-20T21:58:05.752 INFO:teuthology.orchestra.run.plana50.stderr: 1. Contact the upstream for the repository and get them to fix the problem. 2015-08-20T21:58:05.753 INFO:teuthology.orchestra.run.plana50.stderr: 2015-08-20T21:58:05.753 INFO:teuthology.orchestra.run.plana50.stderr: 2. Reconfigure the baseurl/etc. for the repository, to point to a working 2015-08-20T21:58:05.753 INFO:teuthology.orchestra.run.plana50.stderr: upstream. This is most often useful if you are using a newer 2015-08-20T21:58:05.753 INFO:teuthology.orchestra.run.plana50.stderr: distribution release than is supported by the repository (and the 2015-08-20T21:58:05.753 INFO:teuthology.orchestra.run.plana50.stderr: packages for the previous distribution release still work). 2015-08-20T21:58:05.753 INFO:teuthology.orchestra.run.plana50.stderr: 2015-08-20T21:58:05.753 INFO:teuthology.orchestra.run.plana50.stderr: 3. Disable the repository, so yum won't use it by default. Yum will then 2015-08-20T21:58:05.753 INFO:teuthology.orchestra.run.plana50.stderr: just ignore the repository until you permanently enable it again or use 2015-08-20T21:58:05.754 INFO:teuthology.orchestra.run.plana50.stderr: --enablerepo for temporary usage: 2015-08-20T21:58:05.754 INFO:teuthology.orchestra.run.plana50.stderr: 2015-08-20T21:58:05.754 INFO:teuthology.orchestra.run.plana50.stderr: yum-config-manager --disable <repoid> 2015-08-20T21:58:05.754 INFO:teuthology.orchestra.run.plana50.stderr: 2015-08-20T21:58:05.754 INFO:teuthology.orchestra.run.plana50.stderr: 4. Configure the failing repository to be skipped, if it is unavailable. 2015-08-20T21:58:05.754 INFO:teuthology.orchestra.run.plana50.stderr: Note that yum will try to contact the repo. when it runs most commands, 2015-08-20T21:58:05.754 INFO:teuthology.orchestra.run.plana50.stderr: so will have to try and fail each time (and thus. yum will be be much 2015-08-20T21:58:05.754 INFO:teuthology.orchestra.run.plana50.stderr: slower). If it is a very temporary problem though, this is often a nice 2015-08-20T21:58:05.755 INFO:teuthology.orchestra.run.plana50.stderr: compromise: 2015-08-20T21:58:05.755 INFO:teuthology.orchestra.run.plana50.stderr: 2015-08-20T21:58:05.755 INFO:teuthology.orchestra.run.plana50.stderr: yum-config-manager --save --setopt=<repoid>.skip_if_unavailable=true 2015-08-20T21:58:05.755 INFO:teuthology.orchestra.run.plana50.stderr: 2015-08-20T21:58:05.755 INFO:teuthology.orchestra.run.plana50.stderr:Cannot retrieve metalink for repository: epel/x86_64. Please verify its path and try again
http://pulpito.ceph.com/teuthology-2015-08-17_23:08:05-kcephfs-master-testing-basic-multi/1020551/
http://pulpito.ceph.com/teuthology-2015-08-17_23:08:05-kcephfs-master-testing-basic-multi/1020548/
http://pulpito.ceph.com/teuthology-2015-08-17_23:08:05-kcephfs-master-testing-basic-multi/1020533/
http://pulpito.ceph.redhat.com/teuthology-2015-08-16_18:08:02-kcephfs-hammer-testing-basic-magna/175509/
http://pulpito.ceph.com/teuthology-2015-08-21_23:08:01-kcephfs-next-testing-basic-multi/1026122/
Dunno if this is a networking thing or something else, but it popped up in several places.
Updated by Andrew Schoen over 8 years ago
- Status changed from New to In Progress
- Assignee set to Andrew Schoen
I believe this is caused by a timeout to: http://mirrors.fedoraproject.org/metalink?repo=epel-7&arch=x86_64
We initially stared to use mirrorlist because we had timeout issues with dl.fedoraproject.org, https://github.com/ceph/ceph-cm-ansible/pull/117
Changing to the mirrorlist helped us with those timeouts because yum will pick a new mirror if one is failing, but that won't help us if we timeout getting the mirror list from fedoraproject.org.
Perhaps we might want to host that metalink file ourselves?
Updated by Andrew Schoen over 8 years ago
A solution that ktdreyer suggested was to have ceph-cm-ansible create and manage a local mirrorlist on nodes that need it. This way we can have more fine grained control of the mirrors and it will eliminate a point of failure in downloading the file.
Updated by Andrew Schoen over 8 years ago
We have a few directions we could go here:
1) increase the timeout and retires for yum when talking to mirrors
2) Use the mirrorlist link instead of metalink, the thought is that it might timeout less. http://mirrors.fedoraproject.org/mirrorlist?repo=epel-7&arch=x86_64
3) Have cm-ansible manage a local mirror list
The plan is to try option 1 first as it might be less evasive and won't have the maintenance burden of managing our own mirror list.
Updated by Andrew Schoen over 8 years ago
- Status changed from In Progress to Fix Under Review
Updated by Andrew Schoen over 8 years ago
- Status changed from Fix Under Review to Resolved