Epoptes clients disconnect from server randomly

[gdi2k]

Normally, if a user is logged into a machine, epoptes displays an icon of the user's screen and shows user information etc.

However, sometimes the icon turns to the "fat" icon mid-session, without the user having logged out. I understand that this means that the epoptes client has disconnected from the epoptes server.

We get this about 2 - 3 times per day on different clients (no noticeable pattern as to which clients). We have about 35 fat clients. Once it happens, it stays that way - the user's screen never returns.

If I monitor such a machine, the monitoring works fine (VNC window opens showing user's desktop etc.).

I've attached a screenshot showing client 152 (fat icon), it's properties (blank), and the VNC session associated with that client (active user session).

Happy to provide any diagnostic info!

[gdi2k (gdi2k)]

Screenshot.

[Fotis Tsamis (ftsamis)]

Epoptes-client has a logic where if it cannot send a screenshot to the server for 3 times in a row, it assumes that the connection is broken, and exits.

So, if a client's network connectivity is lost for 20 seconds, you won't really notice anything else other than losing the thumbnails etc, the things you've mentioned.

To verify that it's a problem with your network (or not), login as a user, run /usr/sbin/epoptes-client from a terminal, and from another terminal run tail -f /var/log/syslog.

When you lose the thumbnails, check the output of the terminal. I think you'll see a message about "eth0: link is down" or something similar at about the same time that you lost the connectivity with epoptes-client.

[gdi2k (gdi2k)]

Hi Fotis,

Thanks for your comment. I have been monitoring logs and so forth for some time based on your recommendations, but I have seen nothing in syslog to indicate network issues. These are fat clients, so if there were network problems, I feel that nbd would be taken down too and the machine would become unresponsive, which does not happen.

I'll attach a typical syslog for a machine where it has happened today. It has been up 4.5 hours. There's a lot of network manager garbage, and you'll see cron reducing nice levels on skype and linphone a lot, along with teamviewerd crashing a few times. I can't find anything that would affect epoptes though, but hopefully you can.

I don't know the exact time that epoptest-client stopped working in this case.

[Fotis Tsamis (ftsamis)]

Hi gdi2k,

the server syslog doesn't help, it wouldn't catch the temporary network disconnection event of the client.
Please follow the troubleshooting procedure that I described in my previous comment, i.e. run epoptes-client from the command line and check its output. Or/and post the dmesg/syslog of the LTSP client, not of the server.

> I feel that nbd would be taken down too and the machine would become unresponsive, which does not happen.

No, nbd timeout is longer than 17 seconds, and in recent versions it also supports reconnections.
And of course the client doesn't read from NBD all the time as it's caching much of it.

You can

[Launchpad Janitor (janitor)]

[Expired for Epoptes because there has been no activity for 60 days.]