[Gluster-users] random file not founds (enoent) in a distribute+replicate setup

mki-glusterfs at mozone.net mki-glusterfs at mozone.net
Sat Dec 4 00:12:50 UTC 2010


In a distribute+replicate setup, I'm randomly seeing cases where a file 
can't be seen by processes.  A stat of the file using the full path 
results in an ENOENT (No such file or directory).  However if I cd to 
the directory that it's supposed to be in, and then do an ls it appears 
and is fine from there on.

Any ideas as to what might be happening here?  This is 3.0.4 btw, I'm
working to get it updated at the very least to 3.0.6 soonish but need
to get this other cruft cleaned up first, etc.

# ls -al 3/9/0/9/5/263259093
ls: cannot access 3/9/0/9/5/263259093: No such file or directory
# ls -al 3/9/0/9/5
... thousands of files spewed here ...
# ls -al 3/9/0/9/5/263259093
-rw-r--r-- 1 2000 2000 2780068 Nov 12 23:15 3/9/0/9/5/263259093

If a backend server goes offline and comes back on later, would that
cause this issue?  I thought that stat'ing a file would result in
the self heal and eventual consistency across all replicated pairs?
The configs are standard glusterfs-volgen generated raid1 configs
so no read-subvolume directives in there anywhere...



More information about the Gluster-users mailing list