> Additionally, CODA keep crashing. This seems to happen only when you stop
> a run and then wait for about 1 minute or longer before starting a new
> run. When you do so, the visual_scal program freezes and so does the
> EPICS ca_server. You have to shut all the windows and start all over.
> Hopefully, it's just something I'm doing (or not doing), but, if not, it's
> quite annoying...
>
I'm not sure that CODA has crashed, it just seems like it is hung. And
independent of CODA's behavior, I've seen the visual_scaler display
freeze sometimes. Starting a new CODA run seems to free it up.
I've tried to reproduce what Aaron was seeing, where CODA would hang if
you waited too long (>1 minute?) between runs. I took 9 runs, waiting 5
seconds, 1 minute, 3 minutes and >1 hour between runs. I only saw a
hang once (one of the cases where I waited about 1 minute), where
monitoring blastER showed 0 events (even though visual_scal and EPICS
recording were running).
For that run, End Run failed (blastEB not responding). I could see End
Run would time out, so on RunControl I hit Cancel, then Reset, then
Download. Prestart, Go, End worked fine for the next runs. This is
faster than killing all CODA, scal and EPICS windows, and restarting
all. And it appears to be all that is necessary.
BUT you still shouldn't have to do this. Since it only failed once for
me, I haven't been able to see what is wrong. I wasn't running any
physics triggers, since the hall is open for some detector work. If a
hang (no events at all, not just ROC sleep) happens to you:
1) Make sure you are monitoring blastER, not EB or one of the ROCs.
With no physics triggers, EB and ROCs won't count.
2) Look at the EB and ER windows. While a run is in progress, EB should
have reported "Got all fragments of prestart", "Got all fragments of
go". ER should have reported "Got Prestart Event!!", "activating",
"activated". If there is a hang, please cut and paste all the EB screen
text for that run to ~/commis/coda/eb_msg.txt. I've put the text for a
good run in there; maybe something will show up by comparison.
3) End Run (Cancel if you don't want to wait for the timeout), Reset,
Download, then start a new run as usual.
Maybe there are a ton of physics triggers that swamp CODA, because the
wire is too far in? Although in that case, there should still be scaler
and EPICS data being recorded (be sure to monitor ER!); this should be
ROC sleep, and changing the monitoring point to ROC should restore
physics data flow.
Karen
This archive was generated by hypermail 2.1.2 : Mon Feb 24 2014 - 14:07:28 EST