OUT OF MEMORY ERROR KILLED PROCESS....

[pscout]

One other thing i have noticed is the client hanging at the end of a wu. A restart seems to clear it.

When it happened on a previously stable (IIRC ) rig last night, it hadn't yet got a new wu. so instead of just restarting the client, i deleted the work and queue.dat files ... all 9 sets of files for the 9 queue entries were still there. They are supposed to get cleaned up when a wu is xmited but apparantly this is one of the outstanding bugs with smp.

I have no idea if the same scenario can be causing your out of memory symptoms but i wouldn't be suprised. It could be just a different side effect that you see on your rig because of component or config differences.

another common problem with smp seemed to require the use of qfix, but IIRC that was mainly due to not specifying >5MB on the client config. I have yet to need to use qfix ... which is a good thing since i seem to be pretty useless at remembering how to navigate the linux environment with these 2/3 character commands

I don't think this qfix symptom has anything to do with your problem directly, but i suspect it is yet another symptom of the challenges the 4 threaded smp client has in dealing with the file system with integrity.

Also, given how touchy kubuntu seems to be with hardware changes (in my limited experience) , that is the only other guess i could suggest as a posible source ... except of course intermittent hardware issues which i have no clue as to how to diagnose in linux.

 

[CJ-5]

pscout: all 9 sets of files for the 9 queue entries were still there. They are supposed to get cleaned up when a wu is xmited but apparantly this is one of the outstanding bugs with smp

.

This is interesting because I noticed that when I restarted the computer and reinstalled FAH that one of the first things that the log addressed was that some previous wu was past it's deadline.

Each time it would be a different wu all of which had been finished and turned in successfully days or even weeks ago. It's as if all of this is still in memory and not being treated properly by the program.

Maybe it's hoging memory unecessarily. I still don't know what to do about it though or if that's even the real problem here. Mostly I don't understand why others are not experiencing this problem.

 

[pscout]

so far it has happened to you twice? Maybe if it happens a thrid time a pattern will emerge

The biggest problem with diagnosis is usually getting repeatability.

 

[CJ-5]

I think it's about to KILL THE PROCESS again. When I turn on the monitor or try to open another application it takes forever to do anything. I checked the log, something I couldn't do before after the process was killed previously. I found this after each finished wu:

warning: could not delete all work unit files (2) Core returned invalid code

What can I do to fix this? Any ideas?

 

[pscout]

There are still bugs in the smp client with file handling that haven't been fixed.

I have gotten into this habit when a wu ends ... and they sometimes hang there for hours : .... i will first check if the last wu was sent in sucessfully and that there are no wu's left in the queue to transmit. If all are sent, then i delete the work directory (sudo rm -r work) and the queue.dat (rm queue.dat) file to clean all the junk out that seems to get left behind and then the client seems to work fine for a while untill more junk accumulates.

Not sure if this will fix your problem, but it won't hurt .... unless you delete a completed wu like i did before my first coffee on Sat morning : ... now i am more careful about carefully checking the log that all wu's are sent in first

 

[Mustanley]

the "top" command will show you cpu and memory usage in real time of all running/idle processes. That might help you determine if it's really a memory issue.

 

[dropadrop]

You can see how much memory you have free with the command:

free -m

(will show you in megabytes).

 

[CJ-5]

By the time I got up this morning there was no video. I had to restart the computer and after trying to run f@h found the version to be outdated.

I reinstalled the newest SMP beta and it is now running.

I will try all of your helpful suggestions in the near future. Hopefully before things crap out on me again I can figure out how to keep it folding.

Thanks for helping.