Mark,
Here's the patch:
diff --git a/src/condor_schedd.V6/schedd.C b/src/condor_schedd.V6/schedd.C
index 85bfbd7..cc5c03a 100644
--- a/src/condor_schedd.V6/schedd.C
+++ b/src/condor_schedd.V6/schedd.C
@@ -223,6 +223,15 @@ match_rec::~match_rec()
if( pool ) {
free(pool);
}
+ if( request_claim_sock ) {
+ // NOTE: the value passed to Register_DataPtr()
for this
+ // registered socket is just a pointer to this
match_rec,
+ // so there is no need to worry about
deallocating that.
+ daemonCore->Cancel_Socket( request_claim_sock );
+ delete request_claim_sock;
+ request_claim_sock = NULL;
+ scheduler.rescheduleContactQueue();
+ }
}
@@ -11049,16 +11058,6 @@ Scheduler::DelMrec(char const* id)
return -1;
}
- if( rec->request_claim_sock ) {
- // NOTE: the value passed to Register_DataPtr()
for this
- // registered socket is just a pointer to this
match_rec,
- // so there is no need to worry about
deallocating that.
- daemonCore->Cancel_Socket( rec->request_claim_sock );
- delete rec->request_claim_sock;
- rec->request_claim_sock = NULL;
- rescheduleContactQueue();
- }
-
// release the claim on the startd
if( rec->needs_release_claim) {
send_vacate(rec, RELEASE_CLAIM);
--Dan
Mark Calleja wrote:
Dan,
That's great news. While we wait for 7.0.4 to appear, is there a chance
of a patch being released so we can at least build our own fixed
schedd's from source please?
Cheers,
Mark
Dan Bradley wrote:
Mark,
Thanks for the report. I have found the source of trouble. The problem
was introduced in 7.0.1. It affects parrallel and MPI universe jobs on
all platforms. When there is a problem claiming a startd, the schedd
crashes.
This will be fixed in 7.0.4, which we hope to release in the near future
(on the order of weeks, not months).
--Dan
Mark Calleja wrote:
Hi chaps,
We've hit a problem and we'd urgently like to hear of any solution. We
have a Solaris box that acts as a submit host and its schedd dies every
few minutes; the following snippet from the SchedLog is typical of the
symptom:
7/10 10:50:28 (pid:16042) Calling Handler <to startd <172.24.89.88:9108>>
7/10 10:50:28 (pid:16042) ERROR "Assertion ERROR on
(mrec->request_claim_sock == sock)" at line 1361 in file
dedicated_scheduler.C
7/10 10:50:43 (pid:16188)
******************************************************
7/10 10:50:43 (pid:16188) ** condor_schedd (CONDOR_SCHEDD) STARTING UP
7/10 10:50:43 (pid:16188) ** /prg/condor/sbin/condor_schedd
7/10 10:50:43 (pid:16188) ** $CondorVersion: 7.0.3 Jun 20 2008 BuildID:
91405 $
7/10 10:50:43 (pid:16188) ** $CondorPlatform: SUN4X-SOLARIS29 $
7/10 10:50:43 (pid:16188) ** PID = 16188
7/10 10:50:43 (pid:16188) ** Log last touched 7/10 10:50:28
7/10 10:50:43 (pid:16188)
******************************************************
The OS details are:
% uname -a
SunOS <hostname> 5.9 Generic_112233-10 sun4u sparc SUNW,Sun-Fire-880
This seems related to the problem mentioned here:
http://www.cs.wisc.edu/condor/ligo-tickets/2237.html
Was that problem resolved? It's not apparent from the link. For now
we're downgrading that box to 6.8.8, but that can only be a short term
solution
Any clues/fixes out there?
Best regards,
Mark
_______________________________________________
Condor-users mailing list
To unsubscribe, send a message to condor-users-request@xxxxxxxxxxx with a
subject: Unsubscribe
You can also unsubscribe by visiting
https://lists.cs.wisc.edu/mailman/listinfo/condor-users
The archives can be found at:
https://lists.cs.wisc.edu/archive/condor-users/
_______________________________________________
Condor-users mailing list
To unsubscribe, send a message to condor-users-request@xxxxxxxxxxx with a
subject: Unsubscribe
You can also unsubscribe by visiting
https://lists.cs.wisc.edu/mailman/listinfo/condor-users
The archives can be found at:
https://lists.cs.wisc.edu/archive/condor-users/
_______________________________________________
Condor-users mailing list
To unsubscribe, send a message to condor-users-request@xxxxxxxxxxx with a
subject: Unsubscribe
You can also unsubscribe by visiting
https://lists.cs.wisc.edu/mailman/listinfo/condor-users
The archives can be found at:
https://lists.cs.wisc.edu/archive/condor-users/