Track the 'stats' command from the interface... It'll list some useful
information.
Also sit around with !watch open to see if your query workers are dying
for some reason.
Check lsof (or just ls -l /proc/[pid]/fd | wc -l) to see how many fd's
your trackers are using, compared to the default limit of 1024.
Finally, write a small script that'll connect, run some basic get_paths
or something, and disconnect every few seconds. Run one on localhost and
one remotely to see if you can reproduce the issue and if it happens on
localhost as well.
-Dormando
Andrew McClain wrote:
I'm getting a lot of errors on my production servers, saying:
"MogileFSError: MogileFS::Backend: couldn't connect to mogilefsd backend"
They come in spurts, from different requesting IPs, which makes me think
that it's a connectivity/socket overloading issue.
I have two trackers set up, all on the local LAN of our hosting. I just
want to make sure this isn't mogile freaking out before I start grilling
our hosting provider about their inter-machine connectivity.
Are there any stats I can track to help debug this?
Tap the collective.
http://www.fluther.com
