I'm seeing some odd issues mounting a GFS2 filesystem exported via NFS
from a CentOS-7 cluster. The main clients are CentOS-6, but I've also
tested from CentOS-7. There are several filesystem exported from this
host, and not all of them show the problem. At the initial report, these
were the symptoms:
o NFSv4
$ mount -t nfs -o hard,intr $SERVER:/home/socr/a /mnt
$ ls -la /mnt
ls: reading directory /mnt: Too many levels of symbolic links
total 460
drwxr-xr-x. 32 root root 3864 Sep 27 2016 ./
dr-xr-xr-x. 43 root root 4096 Feb 24 10:08 ../
dr-xr-xr-x. 43 root root 4096 Feb 24 10:08 ../
drwxr-xr-x. 59 20383 wheel 2048 Mar 23 12:29 al/
drwxr-xr-x. 59 20383 wheel 2048 Mar 23 12:29 al/
.
.
.
$ dmesg | tail -1
NFS: directory / contains a readdir loop.Please contact your server vendor. The file: .. has duplicate cookie 1
Note that there are *no* symbolic links in that directory or any of the
others that show the issue.
o NFSv3
$ mount -t nfs -o hard,intr,tcp,nfsvers=3 $SERVER:/nfsexports/home/socr/a /mnt
$ ls -la /mnt
.
Here ls hangs, eating up all of a CPU and ever more memory until finally
dying with the message "ls: memory exhausted".
The folks running the server took some of the FSes off line and fsck'd
them. Afterwards, NFSv4 has the same problem (double listing, readdir
loop error in client logs) but NFSv3 doesn't -- ls works, doesn't report
any errors and doesn't double list files and directories.
Our current workaround, obviously, is to fsck the filesystem and use
NFSv3. But that doesn't give me a warm fuzzy feeling. Does this
constellation of symptoms ring any bells? Thanks.
--
Joshua Baker-LePain
QB3 Shared Cluster Sysadmin
UCSF
--
To unsubscribe from this list: send the line "unsubscribe linux-nfs" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
