Message boards : Number crunching : Newest WU not buffering
Message board moderation

To post messages, you must log in.

AuthorMessage
KWSN-GMC-Peeper of the Castle ...
Avatar

Send message
Joined: 6 Oct 05
Posts: 18
Credit: 952,091
RAC: 0
Message 26816 - Posted: 4 Oct 2014, 4:29:42 UTC

Not buffering to RAM or something. Constant disk access like it's pulling data, crunching, writing results, pulling data etc. Causing wild swings in CPU usage.

ID: 26816 · Report as offensive     Reply Quote
Professor Ray

Send message
Joined: 26 Nov 05
Posts: 39
Credit: 49,164
RAC: 10
Message 26817 - Posted: 4 Oct 2014, 4:42:10 UTC

I've noticed the latest batch of W3-'s does a lot of that initially on WU restart. It settles down after a couple of minutes.
ID: 26817 · Report as offensive     Reply Quote
KWSN-GMC-Peeper of the Castle ...
Avatar

Send message
Joined: 6 Oct 05
Posts: 18
Credit: 952,091
RAC: 0
Message 26818 - Posted: 4 Oct 2014, 4:44:02 UTC - in response to Message 26817.  
Last modified: 4 Oct 2014, 4:46:49 UTC

I'm afraid it's a constant problem here. hasn't stopped all evening. I'm a little concerned about how hard it's beating my hdd and if it's going to heat them up too much.

ID: 26818 · Report as offensive     Reply Quote
Professor Ray

Send message
Joined: 26 Nov 05
Posts: 39
Credit: 49,164
RAC: 10
Message 26820 - Posted: 4 Oct 2014, 4:53:37 UTC
Last modified: 4 Oct 2014, 4:58:13 UTC

Is it one of those infamous w3-'s that just appeared in the last coupla-day?

Check you tasks listing on the web-site and check when it was served down to you?

IF it was before now, then edit your client_state.xml to change the rsc_disk parameter for the WU to 600000000 after exiting BOINC manager. See how it runs after that. Again, it'll churn for a few minutes max.

OR you could just abort it. Lots of that going around lately.

You're call: no matter how much time you gots into it, if the disk quota is too low for the WU it'll abend with dik overflow error (unless you make disk quota bigger for WU like I said). If you don't want to edit you client_state.xml and you got any w-3 type WU before now, it(they) will abend end on you anyways.

New WU are being served with disk quota of 1/2 GB per WU, old ones were 190MB. I found at 85% it was at 377MB.

Right now the only WU waiting validation is prolly mine from earlier this afternoon.
ID: 26820 · Report as offensive     Reply Quote
KWSN-GMC-Peeper of the Castle ...
Avatar

Send message
Joined: 6 Oct 05
Posts: 18
Credit: 952,091
RAC: 0
Message 26822 - Posted: 4 Oct 2014, 5:00:16 UTC - in response to Message 26820.  

I guess my stand is if this work isn't important enough to set up the WU properly it's not important enough for me to stress about. I'll just suspend the project for a couple days.

ID: 26822 · Report as offensive     Reply Quote
Professor Ray

Send message
Joined: 26 Nov 05
Posts: 39
Credit: 49,164
RAC: 10
Message 26823 - Posted: 4 Oct 2014, 5:09:04 UTC

You could do that, but resuming in a couple of days won't alter any parameters for WU already downloaded before now.

So if you abort all WU's and suspend the project for a couple of days, you'll be golden in a few days.

Otherwise you'll have to manually edit the way I splained for each WU in client_state.xml; its not a one shot deal for the project.

The way I un'rstan' it: prollem's fixed for WU's being served up now. I'm guessing ones that are constantly streaming back as failed and being served out now, prolly are fixed. Dunno.

I'll take whatever LHC serves me and I'll make 'er crunch 'em. The work being computed is pretty intense - very cool guy stuff - I like it.
ID: 26823 · Report as offensive     Reply Quote
Eric Mcintosh
Volunteer moderator
Project administrator
Project developer
Project tester
Project scientist

Send message
Joined: 12 Jul 11
Posts: 852
Credit: 1,619,050
RAC: 0
Message 26836 - Posted: 4 Oct 2014, 9:47:42 UTC - in response to Message 26823.  

I am trying to figure out how to get rid of these w-b3
WUs now. Maybe Monday before it can be fixed.
ID: 26836 · Report as offensive     Reply Quote
KWSN-GMC-Peeper of the Castle ...
Avatar

Send message
Joined: 6 Oct 05
Posts: 18
Credit: 952,091
RAC: 0
Message 26851 - Posted: 4 Oct 2014, 21:10:39 UTC - in response to Message 26823.  
Last modified: 4 Oct 2014, 21:20:44 UTC

"So if you abort all WU's and suspend the project for a couple of days, you'll be golden in a few days."

yes. of course the existing WU had to go. sorry if I wasn't clear.
No roomful of computers here, doing BOINC on my one, single computer I use for everything else as I always have so I can't have it crippled by constant disk access while I try to do other things.
Make no mistake, the LHC is my favorite project and the parts I can sorta understand without the math are very exciting. That monster has already paid for itself and there's more to come.

ID: 26851 · Report as offensive     Reply Quote
Eric Mcintosh
Volunteer moderator
Project administrator
Project developer
Project tester
Project scientist

Send message
Joined: 12 Jul 11
Posts: 852
Credit: 1,619,050
RAC: 0
Message 26852 - Posted: 4 Oct 2014, 21:51:27 UTC

Well thanks; I think I have sorted it (and learned
a lot). Looks good now, but I shall check in the morning.
ID: 26852 · Report as offensive     Reply Quote
KWSN-GMC-Peeper of the Castle ...
Avatar

Send message
Joined: 6 Oct 05
Posts: 18
Credit: 952,091
RAC: 0
Message 26858 - Posted: 7 Oct 2014, 23:07:05 UTC - in response to Message 26852.  

That seems to be it. Everything looks to be working smoothly. Now we can get back to making the most out of that hardware.

ID: 26858 · Report as offensive     Reply Quote

Message boards : Number crunching : Newest WU not buffering


©2020 CERN