Mantis - Resin
Viewing Issue Advanced Details
ID: Category: Severity: Reproducibility: Date Submitted: Last Update:
20 minor always 03-07-05 00:00 11-30-05 14:43
Reporter: user35 Platform:  
Assigned To: OS:  
Priority: high OS Version:  
Status: closed Product Version:  
Product Build: Resolution: unable to reproduce  
Projection: none      
ETA: none Fixed in Version: 3.0.13  
Summary: 0000020: Sessions don't get destroyed in srun cluster
Description: RSN-11
We are using Resin Professional 3.0.12 with an srun cluster setup. Things work normal for some time, then sessions don't get destroyed any more. Sessions accumulate until the server crashes due to overload. With full logging enabled one can see that the cleanup thread which in normal operation cleans the timeouted sessions every second just diappears at some point in time, hence no more sessions get destroyed. Nothing else in the logs.
Steps To Reproduce:
Additional Information:
Relationships
Attached Files:

Notes
(0000020)
ferg   
03-07-05 00:00   
According to the configuration, there is no persistent store.

So the fact that the configuration is clustered is irrelevant.

Therefore, the maximum number of sessions is the 4096 in the configuration. If that value causes an overload, it would need to be reduced. (Which would indicate that too much information is being put in the session.)
(0000021)
user35   
03-07-05 00:00   
We followed your suggestion to reduce session max (The three in resin.conf defined webapps with new session max values of 2048/200/20), but still the problem occurs.

Some more details:
* some of the nodes work for weeks before the problems occurs ("cleanup" Thread vanishes / does nothing any more)
* on some nodes the problem occurs after some hours
* this happens in random order on all nodes, not always on the same node
* all nodes do the same
(0000022)
ferg   
03-07-05 00:00   
Can't reproduce this issue.