Hi,
We have users who can start an RStudio session and run a job but after a couple of hours their loose the interface and cannot reconnect to it. Any recommendations on where to look? I see not out of memory errors.
Here is an example. Note this happens both via VPN and Citrix. I cannot verify if this also happens on site as we are all working from home. I suspect it does though.
“Just to let you know I get a similar thing with Citrix. I tested by starting an RStudio session with walltime 24 hrs yesterday afternoon, then got disconnected from Citrix overnight. Today when I log back in to Citrix and connect to the running RStudio the environment is now empty (what I had run is gone). In the rsession log I see attached, is it launching a fresh RStudio session instead of connecting to the existing one?
She also reports seeing an error in the output.log. I suspect this is a different issue to the one described above?
Can you expand on that? Does the job end too? I guess the questions we have here is (1) did the process exit before the walltime (and if so why). or (2) Did we just loose the panel in the OOD interface and it’s actually still running in the batch system. or (3) job is still running, panel still exists in the OOD UI and users just can’t click through (rstudio returns some error).
For issue 1, you may want to add a thing at the end of the script to capture the commands exist status and print it out. Something like this.
singularity run ...
STATUS=$?
echo "singularity rserver exited with status $STATUS at $(date)"
As to the other issue, it could be related. Maybe the rsession is ending so the rstudio server stops? I’m not sure, I don’t run these apps for that long, so it’ll take me a while to replicate.
The only way I could replicate your second issue is to force the session to end. There’s a power button at the top right to do that. But note, Even if I ‘logged out’ (the button right next to the power button) It wouldn’t restart my rsession.
I don’t think VPN or Citrix has much to do with it. It may have something to due with RStudio’s session timeout configuration. Or the user may have just shut down the rsession themselves by accident. I tried pretty hard to replicate it happening without my forcing it and I couldn’t. I had a job running for 17 hours and all.
As to your first issue, let me know some details and we’ll figure it out.I tested with versions 1.2.5001 and 1.1.380.
Thx mate. Back to u in a couple of days.
regards,
Christopher Welsh
On his mobile.
Jeff, turns out to be the session timeout. Sorry for the delay getting back to u. No one final thing preventing us from going live is that some valid users are not recognised when they try login. I will raise a new thread. Thx.
