Hi, Our first attempt at using CephFS in earnest in December ran into a known bug with the kclients hanging in ceph_mdsc_do_request, which I suspected was down to the bug in http://comments.gmane.org/gmane.comp.file-systems.ceph.devel/15838. We were on a default Ubuntu 12.04 3.2 kernel, so recently we switched to the saucy 3.11 kernel, which I see included the patch to fix it (3.11.0-18-generic #32~precise1-Ubuntu). We are using ceph version 0.72.2 (a913ded2ff138aefb8cb84d347d72164099cfd60).
Things work much better now, but occasionally after hours of heavy load the kclient seems to have a stale view of the filesystem, as if the cached directory is not being updated. I'm not sure if it's related, but on a machine this happened to, I couldn't read the dentry cache :- root@xxx:/sys/kernel/debug/ceph/e7cc20a1-316f-447f-962f-73a2a535b121.client4472# cat dentry_lru cat: dentry_lru: Cannot allocate memory Remounting the filesystem fixed these issues, but obviously this makes it unuseable currently. I have a couple of questions :- 1. When this happens, what further debugging can I do on the client side? The mds connection still looks active, and there are no funny messages in the kernel log. 2. Is there a single place what I can see all cephfs patches, and how can I tell which kernels they are being applied to? Or if I am keen on being on the CephFS bleeding edge, should I just maintain my own? CephFS would solve a lot of our storage problems, so I'm keen on helping get it stable and supported. Best regards, Danny
_______________________________________________ ceph-users mailing list ceph-users@lists.ceph.com http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
