Panic Mode On (110) Server Problems?

Message boards : Number crunching : Panic Mode On (110) Server Problems?
Message board moderation

To post messages, you must log in.

Previous · 1 · 2 · 3 · 4 · 5 . . . 37 · Next

AuthorMessage
Profile Keith Myers Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 13164
Credit: 1,160,866,277
RAC: 1,873
United States
Message 1914339 - Posted: 21 Jan 2018, 2:47:38 UTC - in response to Message 1914337.  

I've never seen any recommendation for the special app to use anything other than the -nobs no blocking sync flag. The whole reason for the special app is to utilize the maximum potential of the graphics card. Not sure where the -bs blocking sync flag would be useful. Maybe the low end dual core and 1050 class systems?

Check out the title of thread.
Linux CUDA 'Special' App finally available, featuring Low CPU use
Low CPU use was it's big selling point.
:-)

Hey Grant, but you know we high credit wannabees just skipped right past that part of the thread title and glossed over the low CPU use part. We focussed on the 'Special' part. :-}
Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)
ID: 1914339 · Report as offensive
Profile Jeff Buck Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester

Send message
Joined: 11 Feb 00
Posts: 1441
Credit: 148,764,870
RAC: 0
United States
Message 1914355 - Posted: 21 Jan 2018, 3:34:24 UTC - in response to Message 1914332.  

I've never seen any recommendation for the special app to use anything other than the -nobs no blocking sync flag. The whole reason for the special app is to utilize the maximum potential of the graphics card. Not sure where the -bs blocking sync flag would be useful. Maybe the low end dual core and 1050 class systems?
As Grant mentioned, the original publicly available Special App version that launched the Linux thread had Blocking Sync built in. I ran into a problem with that on the first machine I tried it on, seemingly slowing down one of the GPUs that was on a riser cable off an x8(x4) slot. That caused me to switch to Petri's version without Blocking Sync, although the other two GPUs in that machine worked fine either way.

My other two Linux machines have been tested both ways and, reluctant as I am to lose a full core to support each GPU, I ultimately ended up using -nobs across the board. I think my last post on the subject was back in July (Message 1878599 ), in response to a question you had, as a matter of fact. :^)
ID: 1914355 · Report as offensive
Profile Keith Myers Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 13164
Credit: 1,160,866,277
RAC: 1,873
United States
Message 1914370 - Posted: 21 Jan 2018, 4:31:11 UTC - in response to Message 1914355.  
Last modified: 21 Jan 2018, 4:32:10 UTC

Thanks for the link, Jeff. Too long ago to remember any of that conversation. I don't believe I ever ran with Blocking Sync on. I figure I have enough cores on the Ryzen to support my 3 gpu tasks.

[Edit] I came late to the party of the special app and missed the original release.
Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)
ID: 1914370 · Report as offensive
Profile KWSN THE Holy Hand Grenade!
Volunteer tester
Avatar

Send message
Joined: 20 Dec 05
Posts: 3187
Credit: 57,163,290
RAC: 0
United States
Message 1914447 - Posted: 21 Jan 2018, 16:59:57 UTC

getting back to the thread's actual purpose - I think one or more of the splitters need(s) another kick... 77,000 WU's available to send, but the splitters are only spiting out 26 WU/sec...
.

Hello, from Albany, CA!...
ID: 1914447 · Report as offensive
juan BFP Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 16 Mar 07
Posts: 9786
Credit: 572,710,851
RAC: 3,799
Panama
Message 1914448 - Posted: 21 Jan 2018, 17:08:14 UTC
Last modified: 21 Jan 2018, 17:08:25 UTC

I can't understand why the splitters always needs a kick. Where is the watchdog?
ID: 1914448 · Report as offensive
Richard Haselgrove Project Donor
Volunteer tester

Send message
Joined: 4 Jul 99
Posts: 14679
Credit: 200,643,578
RAC: 874
United Kingdom
Message 1914451 - Posted: 21 Jan 2018, 17:18:34 UTC - in response to Message 1914448.  

I can't understand why the splitters always needs a kick. Where is the watchdog?
Probably at home with her puppies, on a Sunday morning ;-)
ID: 1914451 · Report as offensive
Stephen "Heretic" Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 20 Sep 12
Posts: 5557
Credit: 192,787,363
RAC: 628
Australia
Message 1914471 - Posted: 21 Jan 2018, 20:41:37 UTC - in response to Message 1914318.  
Last modified: 21 Jan 2018, 20:56:38 UTC

[quote]as well as 1CPU core (thread) for each GPU WU being processed to keep those GPUs fed


Is this necessary with the low CPU CUDA80 apps?


. . That depends on whether you run with blocking sync on or off. With it on then you can get away with just the one free CPU core per 2 or 3 GPU tasks, with it off you will need 1 CPU core per GPU task plus one more CPU core free. With all the high thread count CPUs out there that should not be a problem. Even on my core 2 Duo that means no crunching at all on the CPU, but it works well. This is much the same as having sleep on or off running SoG. It is purely up to you, but there is a speed gain with it off.

Stephen

. .
ID: 1914471 · Report as offensive
Stephen "Heretic" Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 20 Sep 12
Posts: 5557
Credit: 192,787,363
RAC: 628
Australia
Message 1914474 - Posted: 21 Jan 2018, 20:44:31 UTC - in response to Message 1914324.  

? ? ? ? The CUDA80/90 apps need and use 1 full cpu core to support the gpu task.

I thought he default was to use Blocking Sync to keep the CPU load down? Disabling blocking sync bumps up the CPU requirements, but also bumps up the output from the video card.


. . Correct ...

Stephen

:)
ID: 1914474 · Report as offensive
Stephen "Heretic" Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 20 Sep 12
Posts: 5557
Credit: 192,787,363
RAC: 628
Australia
Message 1914475 - Posted: 21 Jan 2018, 20:51:00 UTC - in response to Message 1914332.  

I've never seen any recommendation for the special app to use anything other than the -nobs no blocking sync flag. The whole reason for the special app is to utilize the maximum potential of the graphics card. Not sure where the -bs blocking sync flag would be useful. Maybe the low end dual core and 1050 class systems?


. . Hi Keith,

. . TBar went to a lot of trouble to make Blocking Sync the default so the app would be easy to use on "rank and file" crunchers without tweaking. You do not need to use a -bs option due to this effort, it IS the default. And it is useful, not because of low end GPUs, but because of low end CPUs that may not handle the load. With serious CPU power and loads of cores then I would say go -nobs all the way. As Grant said, the GPU performance boost from this sacrifice will be greater than the productivity of the core you are removing from crunch duty.

Stephen

:)
ID: 1914475 · Report as offensive
Grant (SSSF)
Volunteer tester

Send message
Joined: 19 Aug 99
Posts: 13855
Credit: 208,696,464
RAC: 304
Australia
Message 1914542 - Posted: 22 Jan 2018, 5:42:12 UTC - in response to Message 1914448.  

I can't understand why the splitters always needs a kick. Where is the watchdog?

The problem is the servers have reached their limit.

Lately the output of the splitters is limited by the number of MB WU-awaiting-deletion. It's a combination of those, and the received-last-hour and In-progress numbers.
Generally, when the WU-awaiting-deletion backlog clears, the spliters will crank out the work. However that activity slows down the wu-awaiting-deletion processing. So as that backlog increases it eventually gets to a point when the splitter output falls back down. And it stays there (or even drops further) until such time as the WU deleters catch up.
Grant
Darwin NT
ID: 1914542 · Report as offensive
Profile Keith Myers Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 13164
Credit: 1,160,866,277
RAC: 1,873
United States
Message 1914544 - Posted: 22 Jan 2018, 6:20:02 UTC

Looks like DNS issues again. Can't get any work from the project.
Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)
ID: 1914544 · Report as offensive
Grant (SSSF)
Volunteer tester

Send message
Joined: 19 Aug 99
Posts: 13855
Credit: 208,696,464
RAC: 304
Australia
Message 1914545 - Posted: 22 Jan 2018, 6:52:56 UTC - in response to Message 1914544.  

Looks like DNS issues again. Can't get any work from the project.

Around this time of the day, every day (except for yesterday for some reason?) The forums & Scheduler take some random time off. Anything from 10min to 45min, usually around 20min. It usually shows up on the graphs as a drop in work returned, then a surge, then back to what it was. Today it's shown up as a gap in the graphs.
Just picked up 47 WUs after finally making contact with the Scheduler again.
Grant
Darwin NT
ID: 1914545 · Report as offensive
Grant (SSSF)
Volunteer tester

Send message
Joined: 19 Aug 99
Posts: 13855
Credit: 208,696,464
RAC: 304
Australia
Message 1914551 - Posted: 22 Jan 2018, 7:44:45 UTC

Hope they load up some more data soon- we've just about finished off the current batch of work.
Grant
Darwin NT
ID: 1914551 · Report as offensive
Profile Keith Myers Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 13164
Credit: 1,160,866,277
RAC: 1,873
United States
Message 1914553 - Posted: 22 Jan 2018, 8:18:51 UTC - in response to Message 1914545.  

Yes, I've noticed that gap in the graphs everyday. Just hadn't connected it the general sluggishness of the site and connections never going through and timing out. Must not have been online at that time or something.
Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)
ID: 1914553 · Report as offensive
rob smith Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer moderator
Volunteer tester

Send message
Joined: 7 Mar 03
Posts: 22541
Credit: 416,307,556
RAC: 380
United Kingdom
Message 1914554 - Posted: 22 Jan 2018, 8:26:22 UTC

Timing is coincidental with UK breakfast time, so do servers like tea and soft boiled eggs?
or is it Marmite soldiers???
Bob Smith
Member of Seti PIPPS (Pluto is a Planet Protest Society)
Somewhere in the (un)known Universe?
ID: 1914554 · Report as offensive
Profile Dr.Diesel Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor

Send message
Joined: 14 May 99
Posts: 41
Credit: 123,695,755
RAC: 139
United States
Message 1914576 - Posted: 22 Jan 2018, 13:52:06 UTC - in response to Message 1914471.  

[quote]as well as 1CPU core (thread) for each GPU WU being processed to keep those GPUs fed


Is this necessary with the low CPU CUDA80 apps?


. . That depends on whether you run with blocking sync on or off. With it on then you can get away with just the one free CPU core per 2 or 3 GPU tasks, with it off you will need 1 CPU core per GPU task plus one more CPU core free. With all the high thread count CPUs out there that should not be a problem. Even on my core 2 Duo that means no crunching at all on the CPU, but it works well. This is much the same as having sleep on or off running SoG. It is purely up to you, but there is a speed gain with it off.

Stephen

. .


Thanks to all for the help! Wish there was a wiki for this info!

I applied for an account over at Crunchers Anonymous a few weeks ago, but so far nobody has approved my request. :(
ID: 1914576 · Report as offensive
Profile Keith Myers Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 13164
Credit: 1,160,866,277
RAC: 1,873
United States
Message 1914597 - Posted: 22 Jan 2018, 16:50:39 UTC - in response to Message 1914576.  

I believe you only need to register if you want to post in the forums. I think you still have access to read posts and download. But I could be wrong. Been a while since I registered.

Maybe a kindly nudge to Arkayn here via PM might get your registration looked at. I noticed he got around to doing some work here lately so he can post again here at Seti.
Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)
ID: 1914597 · Report as offensive
Stephen "Heretic" Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 20 Sep 12
Posts: 5557
Credit: 192,787,363
RAC: 628
Australia
Message 1914642 - Posted: 22 Jan 2018, 21:14:14 UTC - in response to Message 1914597.  

I believe you only need to register if you want to post in the forums. I think you still have access to read posts and download. But I could be wrong. Been a while since I registered.

Maybe a kindly nudge to Arkayn here via PM might get your registration looked at. I noticed he got around to doing some work here lately so he can post again here at Seti.


. . I got the impression that was only so that he could create the new Panic: Server problems thread when Bernie closed off the old one :)

Stephen

??
ID: 1914642 · Report as offensive
Profile Mr. Kevvy Crowdfunding Project Donor*Special Project $250 donor
Volunteer moderator
Volunteer tester
Avatar

Send message
Joined: 15 May 99
Posts: 3807
Credit: 1,114,826,392
RAC: 3,319
Canada
Message 1914644 - Posted: 22 Jan 2018, 21:17:57 UTC - in response to Message 1914597.  

I believe you only need to register if you want to post in the forums.


RAC of >=1 is also required to post in any of the Message Boards forums, but not in the Questions and Answers forums so people having technical issues getting started can ask for help (so of course we do get spammers there.)
ID: 1914644 · Report as offensive
Profile arkayn
Volunteer tester
Avatar

Send message
Joined: 14 May 99
Posts: 4438
Credit: 55,006,323
RAC: 0
United States
Message 1914650 - Posted: 22 Jan 2018, 22:19:13 UTC - in response to Message 1914576.  

[quote]as well as 1CPU core (thread) for each GPU WU being processed to keep those GPUs fed


Is this necessary with the low CPU CUDA80 apps?


. . That depends on whether you run with blocking sync on or off. With it on then you can get away with just the one free CPU core per 2 or 3 GPU tasks, with it off you will need 1 CPU core per GPU task plus one more CPU core free. With all the high thread count CPUs out there that should not be a problem. Even on my core 2 Duo that means no crunching at all on the CPU, but it works well. This is much the same as having sleep on or off running SoG. It is purely up to you, but there is a speed gain with it off.

Stephen

. .


Thanks to all for the help! Wish there was a wiki for this info!

I applied for an account over at Crunchers Anonymous a few weeks ago, but so far nobody has approved my request. :(


I approved the account fairly soon after you created the account.

ID: 1914650 · Report as offensive
Previous · 1 · 2 · 3 · 4 · 5 . . . 37 · Next

Message boards : Number crunching : Panic Mode On (110) Server Problems?


 
©2024 University of California
 
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.