LLRnet server rally port 300 May 23rd-25th

[Brucifer]

I had to reboot all my systems to clear them out. I am now running the PG app on all but 4 cores. On my quad I have the cach set at 100 each, and at 50 on the X2's. That should significantly reduce the loading on the server hopefully.

edit: and yes, a massive waste of cpu cycles. :( As for switching to another server at this point it would be a slow process to switch over unless I just dumped all the caches. I assume though that cancelling them out would allow them to be automatically recycled in the server queue???

[em99010pepe]

I suppose those pairs would be received by the new server as rejected ones but still valid.

[Brucifer]

Originally Posted by em99010pepe

I suppose those pairs would be received by the new server as rejected ones but still valid.

???????

I was talking about cancelling them out with the issuing server, not just switching to a new server.

[gd_barnes]

Originally Posted by Brucifer

I had to reboot all my systems to clear them out. I am now running the PG app on all but 4 cores. On my quad I have the cach set at 100 each, and at 50 on the X2's. That should significantly reduce the loading on the server hopefully.

edit: and yes, a massive waste of cpu cycles. :( As for switching to another server at this point it would be a slow process to switch over unless I just dumped all the caches. I assume though that cancelling them out would allow them to be automatically recycled in the server queue???

OK, we have rock-and-roll!! I didn't have to reboot my slower cores but I DID have to go into task-manager and do 'end process'. It's strange that hitting 'kill' in the GUI didn't 'clear them out'.

2 slower cores now running. Going after the rest now...


Gary

[glennpat]

I had 3 cores running each of a seperate box a few hours ago when it had got stuck. I did not have to restart mine. They just start working again when what ever was fixed. I was running the regular LLR on windows XP.

[gd_barnes]

Everyone, if LLRnet port 300 is sleeping, i.e. not running, on your machines, do an 'end process' in task manager and then restart it. That seems to do the trick.

I'll cross my fingers that getting my 16 more cores running it won't cause any problems.

[mdettweiler]

Originally Posted by Brucifer

???????

I was talking about cancelling them out with the issuing server, not just switching to a new server.

In that case, you'd want to either use the PrimeGrid client's CancelJob function in llr-clientconfig.txt, or by repeatedly running LLRnet with the -c option (./llrnet -c) until it says "no more job to cancel".

[mdettweiler]

Originally Posted by gd_barnes

Everyone, if LLRnet port 300 is sleeping, i.e. not running, on your machines, do an 'end process' in task manager and then restart it. That seems to do the trick.

I'll cross my fingers that getting my 16 more cores running it won't cause any problems.

From what I'm seeing, the server seems to be all back to normal again--and as long as everyone's got their clients fixed, we're all set and shouldn't need a second server. (Not to mention that setting up a second server would potentially create a major headache as far as stats are concerned--and k/n pairs crunched on the second server wouldn't be counted in the stats at http://nplb.rieselprime.org.)

Now that that's all cleared up, let's get on with the rally and find some primes!

[Brucifer]

Originally Posted by Anonymous

In that case, you'd want to either use the PrimeGrid client's CancelJob function in llr-clientconfig.txt, or by repeatedly running LLRnet with the -c option (./llrnet -c) until it says "no more job to cancel".

yup, that's what I was talking about.

[Brucifer]

Originally Posted by glennpat

I had 3 cores running each of a seperate box a few hours ago when it had got stuck. I did not have to restart mine. They just start working again when what ever was fixed. I was running the regular LLR on windows XP.

Unfortunately the linux one doesn't behave that way. It really hangs nasty, and after trying to kill the process, it was easier to just reboot the systems.

[em99010pepe]

Originally Posted by Brucifer

???????

I was talking about cancelling them out with the issuing server, not just switching to a new server.

Ok but I was referring to the fact that even moving to a new server the work would be received as valid but stored in a rejected txt file. Anyway, let's crunch!