Bug #7429
closed
fedora 19 VMs: leveldb must be upgraded to something newer than 1.9.0
Added by Joao Eduardo Luis about 10 years ago.
Updated about 10 years ago.
Description
When deploying monitors on fedora 19 VMs, the monitors fail to mkfs due to some segfault. This happens regardless of being deployed via ceph-deploy or manually.
This behavior goes away once leveldb is upgraded to latest 1.12.
leveldb 1.9 is also over 1 year old and appears to be installed by default on the vms. It looks like the ceph package lists leveldb as a dependency, but does not enforce a version (if that's even possible).
- Assignee set to Sandon Van Ness
- Priority changed from High to Immediate
All Fedora tests are still failing. We need to address this as soon as possible. I am not sure what entails upgrading leveldb on those distros.
On the vm I deployed when tracking this, it only took me a 'yum upgrade leveldb' or something similar. I would think the master VM would need to be updated in a similar manner?
- % Done changed from 0 to 80
This should be fixed in ceph-qa-chef. I will run a new ceph-deploy run on vps and see how it goes.
- % Done changed from 80 to 100
Ok so this is done but ceph-deploy tests are still failing...
Also that being said it looks like centos/rhel are running leveldb 1.7.0 so are they backporting stuff? I am kind of confused why only fedora needed a newer version.
In my standup Ian talked to me about updating the spec so it requires a newer version of leveldb but I am not sure that makes sense if rhel/centos don't have the newer version.
- Status changed from New to Resolved
So I did a lot of testing on Fedora today. This is done so closing it out.
It looks like fedora18 was now kernel panicking when ceph-deploy commands were running. I was able to get around this by using the newest kernel distro. The problem with running the suite with the new kernel distro is that all the debian runs fail when trying to enable kdb. I guess I could have it not attempt that if the machine is a vps as we don't have serial console setup anyway.
Even then the tests are still failing but now its health status is not OK after 15 minutes.
Also available in: Atom
PDF