AnsweredAssumed Answered

Custom Probes and Controller: Port dropped

Question asked by cymen.vig on Nov 5, 2008
Latest reply on Nov 19, 2008 by cymen.vig
A custom probe written in house is occasionally restarting during a run (and no exit message is printed). The probe typically completes a run in about 55 seconds and is written in Perl using the NimBUS SDK. I noticed in the log of the controller probe an entry is printed at approximately the restart time like so:

Nov  4 11:11:31:501  Controller: Port dropped: probe_name 48011

What is the best way to work around this? How often does the controller probe call the _status callback of each probe?

A few more details:
 - we are using a session
 - we use the _timeout callback for handling scheduling so callbacks are useful
 - the probe runs approximately every 5 minutes so there is a 25s/300s = 1/12 = 8.33% chance the controller will not get a response in 30 seconds from the _status call to the probe and restart it

Ideas considered:
 - threading so child is busy but parent can respond to _status but no threading support in NimBUS perl
 - do only so much work every time the probe is called with _timeout -- so set timeout to 30s, do about 15s of work and then have open window for _status calls - considered but would prefer to avoid complicated approach if some basic solution would work

10/10/2008: I opened a support case asking for more details on this. I'll post back with an update.

Outcomes