Last modified: 2013-05-20 08:14:18 UTC
GlusterFS has caused too much troubles since we started the project. We need to get rid of it entirely. Related mail a week ago http://lists.wikimedia.org/pipermail/labs-l/2013-April/001143.html
Coren has been kind enough to give me the parameters to connect to the NFS server. I will rsync the files from GlusterFS and get the puppet changes needed to point to the new mount point (/srv/project/)
Hashar: Dupe of bug 46823? If not I'd love to understand the difference.
(In reply to comment #2) > Hashar: Dupe of bug 46823? If not I'd love to understand the difference. This bug is for the beta project. I am going to migrate it to use the NFS Server that has been created for the 'tools' project. Beta is going to be the second project to use that server, I guess once it works fine that NFS server will replace GlusterFS and thus solve bug 46823.
bastion apache32 and apache33 have been migrated thanks to Mark A. Pelletier. The migration is handled by applying the role::labsnfs::client class and rebooting the instance. Once done, /data/project points to the labs NFS server instead of GlusterFS. The page went from 560ms to 260ms. I have pinged labs-l about it. Still have to migrate the rest of the instances though.
There is no way to get rid of Gluster so one will have to apply the role::labnfs::client class. I will clean up the old Gluster /data/project to make sure nothing is using it.