From: Jonathan Quick <jon_at_email.protected>
Date: Thu, 26 Mar 2015 16:23:48 +0200

Hi Stuart

On Thu, March 26, 2015 10:35 am, Stuart Weston wrote:
> We are having issues with the schedule:
>
> The 30m is running the new FS on debian wheezy :
>
> 2015.085.01:31:21.44;Log Opened: Mark IV Field System Version 9.11.6
>
> It seems some of the "disk2files" are trying to save more than the 1s ! This
> then means that the next mk5 command fails and the "mk5cn" times out. After
> that the schedule goes wrong, all mk5 commands time out.
>
> We have to stop "DIMino" and restart it. The schedule then catches up and
> continues.
>
> I'll pass this onto Ed, but might want to stop all the "disk2files" in the
> rest of the schedules. Or check the time extraction syntax.

I suspect if you look closely that the disk2file is probably asking for the
right slice but DIMino is then not extracting it correctly. The code in
jive5ab is a lot more reliable in that regard.

The FS should be including 'disk2file=abort' commands to protect against this,
but the Linux OS misbehaves if you allow uncontrolled asynchronous buffered
disk writes as that fills up system memory if the disk falls behind. The
jive5ab code has a limits on the buffering to prevent this. We needed this
fix to be able to play the near-realtime correlation game with Tsukuba.

Regards
  Jon
Received on 2015-03-27 01:24:07