Hi, thanks for you reply. En/na Krishna Srinivas ha escrit:
Jordi, With the information you have given it is difficult to guess what might be causing the problem. The "Connection refused" message indicates that the server process was not running. Can you check?
Yes, i've checked, the glusterd processes on node's site were running all the time. I've got a cronjob which checks that every minute, and there was no report of problems for that matter. A part from that, as i said in my previous message, the rest of client mounting gluster didn't notice anything, everything kept working fine, they all saw the data, so... it doesn't look like there was any kind of problem on glusterfs itself, but on the client mounting it. And that's why i didn't realise about that error message in the log file after several days. I kept the cron checking if the gluster deamons were running and the filesystem available, but noone was watching the log files.
About the stale mount point - were the commands hanging when tried to operate on the mount point? and never returned? or were the giving errors? like "Transport endpoint not connected?
Yes, the commands were hanging. Actually, i had to kill the processes which were accessing gluster. No error message was shown.
1.3.x releases are old, you could try 2.0 rc1
Well... i wish i could, but it's been in production for some time now, and i don't think it's an easy task to migrate from one version to another without shutting everything down. I've read in your website that "2.0 is backward compatible with v1.x disk layout". Does that mean i can upgrade without shutting the system down? However, i'm not saying just no.... if anyone can throw some light on this matter and knows how to make the transition quick and safe... i will be very pleased to know about it.
Using six subvolumes for ns-afr is not a good idea as it will have an unneeded performance hit. You could use two subvols or maybe three if you are paranoid.
Thanks for the tip, i'll take that into consideration and rearrange the conf files so that they split namespaces in different subvolumes.
So... to sum up.... glusterfs has been working for months without problems. I haven't noticed any performance or stability-related issue. However... without human intervention of any kind, it started to log that error message wich worries me and eventually, one client failed to access the filesystem properly. Just let me know if you need any further information about this issue, i've got all the logs generated with WARNING level.
Thanks.
