Hide Forgot
Thu Dec 11 23:39:32 2008 vikas - Ticket created Two users have reported this: A simple client-side AFR configuration. At some point the permissions on the mountpoint reportedly change to 666 from 755. Haven't been able to reproduce yet. -------------------------------------------------------------------------------- # Thu Dec 11 23:56:39 2008 vikas - Comments added ------ message from Elbert Lai to gluster-users -------- Hi, We've been using an AFR configuration of gluster where there are a few gluster servers and many gluster clients. We've noticed that recently, the permissions on the mount point from the client's perspective occasionally change to 444. There's no evidence of sudo or chmod activity in system logs, and cron.d, cron.daily, logrotate.d all come up clean as well, which is a good thing. We've been trying to track down any rogue scripts, but it's worth checking out if there are circumstances in which gluster might change the permissions of its mountpoint. When the permissions do change to 444 on the mountpoint, they also change on /var/gluster/ns to match (as expected). Anyone seen this on a gluster environment before? -------------------------------------------------------------------------------- # Thu Dec 11 23:57:55 2008 vikas - Comments added ------ Seen on IRC --------- < clauc> I'm having a permission issue with the mount point. I'm using client side AFR for replicating the document root of a website accross 3 nodes, but after a while from mounting the glusterfs to the mount point directory, the perissions of the mount point change from 755 to 644, so the apache user can no longer access it and I get permission denied when trying to access the website < clauc> if I change the permissions of the mount point back to 755 it works only untill they are reset again to 644 < clauc> any ideas what may be causing this? -------------------------------------------------------------------------------- # Mon Mar 23 17:54:12 2009 amar - Correspondence added Any update on this? fix done? outstanding? please give an update. -- Amar Tumballi -------------------------------------------------------------------------------- On Mon Mar 23 17:54:12 2009, amar wrote: > Any update on this? fix done? outstanding? please give an update. One more user has reported this. However, he too does not know how to reproduce. We have never encountered this in our own testing. So I suggest we keep this open (or mark as stalled).
We see this too, though it is rare. We have 2 bricks, and 15 or so clients mirroring across them. We're using version 1.3.12 Our mount point loses all it's execute bits - obviously occurs on all the mounts on all the clients. This usually happens around the time we get a stuck client process after concurrent deletes. (see bz #112).
closing the ticket as its of pre-2.0 time. Currently we are focusing on only 3.0.0 releases. Please re-open when see with 3.0.0rc+ versions (rc not made yet). Marking as "WONTFIX"