Panic Mode On (82) Server Problems?


log in

Advanced search

Message boards : Number crunching : Panic Mode On (82) Server Problems?

Previous · 1 · 2 · 3 · 4 · 5 . . . 24 · Next
Author Message
Horacio
Send message
Joined: 14 Jan 00
Posts: 536
Credit: 75,823,032
RAC: 23,057
Argentina
Message 1340906 - Posted: 26 Feb 2013, 0:24:52 UTC

I wish I were able to fully understand why when using a proxy everything runs perfectly but not without it... Well, I wouldn care about that if the owners of the proxies were not filtering my traffic after a couple of days/hours :b
____________

Profile Michael W.F. Miles
Avatar
Send message
Joined: 24 Mar 07
Posts: 249
Credit: 29,164,026
RAC: 7,656
Canada
Message 1340911 - Posted: 26 Feb 2013, 0:33:49 UTC - in response to Message 1340830.

147.31.182.137:80 is working very well
Thanks Bro

Michael Miles

Cosmic_Ocean
Avatar
Send message
Joined: 23 Dec 00
Posts: 2326
Credit: 8,867,769
RAC: 976
United States
Message 1340916 - Posted: 26 Feb 2013, 0:53:25 UTC - in response to Message 1340906.

I wish I were able to fully understand why when using a proxy everything runs perfectly but not without it... Well, I wouldn care about that if the owners of the proxies were not filtering my traffic after a couple of days/hours :b

I believe it has something to do with packet re-try counters/durations, or something along those lines. Maybe a proxy ends up having a smaller MTU on the output side, or for a TCP transfer, instead of 60 seconds before saying "I didn't receive X packet," it is more like 10 seconds, and therefore it can keep pushing it along before BOINC decides it is a stalled transfer and gives up.

It's hard telling. I don't know why exactly proxies seem to do the trick, but the bottom line is that they work wonders.....once you find one that works.
____________

Linux laptop uptime: 1484d 22h 42m
Ended due to UPS failure, found 14 hours after the fact

Tom*
Send message
Joined: 12 Aug 11
Posts: 114
Credit: 4,815,461
RAC: 82
United States
Message 1340923 - Posted: 26 Feb 2013, 1:44:41 UTC
Last modified: 26 Feb 2013, 1:49:54 UTC

Someone reported here back in 2012 that they were fed up with Proxies
they were using failing to work after a few days for their Windows machines and created a local LINUX box running a Proxy server

and had all their Windows Boxes use that local Proxy and it worked.

I have mentioned several times that until we can get co-located the powers that be should implement their own Proxy we can use to talk up the hill.

This would accomplish several things, concentrate the traffic so control messages like acks naks etc do not get lost, also we would not be using resources we do not pay for. Don't they have a couple of boxes retired after upgrades?

*nix boxes seem to talk much more reliably with SETI I would hazard a guess that
99% of Proxy machines use *nix

Profile TRuEQ & TuVaLu
Volunteer tester
Avatar
Send message
Joined: 4 Oct 99
Posts: 479
Credit: 19,963,078
RAC: 19,958
Sweden
Message 1340935 - Posted: 26 Feb 2013, 5:06:06 UTC
Last modified: 26 Feb 2013, 5:06:27 UTC

More then 24hours for 5ap dl now.
But they will leave the MEM soon, they are 60-90% dl now.

server/router pipes must be clogged with left behind transfers like mine. Data just sits there waiting to be sent.

Profile TRuEQ & TuVaLu
Volunteer tester
Avatar
Send message
Joined: 4 Oct 99
Posts: 479
Credit: 19,963,078
RAC: 19,958
Sweden
Message 1340936 - Posted: 26 Feb 2013, 5:08:38 UTC

And why is this allowed?

<exclude_gpu>
<url>http://setiathome.berkeley.edu</url>
<device_num>1</device_num>
</exclude_gpu>
<max_file_xfers>12</max_file_xfers>
<max_file_xfers_per_project>4</max_file_xfers_per_project>
</options>
</cc_config>

There's no limit for how many pipes I can use for dl with this option.

Maybe limit to 1 pipe each could help??

Profile BMaytum
Volunteer tester
Avatar
Send message
Joined: 3 Apr 99
Posts: 101
Credit: 3,974,014
RAC: 3,526
United States
Message 1340938 - Posted: 26 Feb 2013, 5:30:01 UTC

Well 9-10 days ago, the U/Ls, D/Ls & Reporting throughput was awesomely good, for example: http://setiathome.berkeley.edu/forum_thread.php?id=70730&postid=1338911.

Then S@H had the weekend power maintenance outage, and in the 1-1/2 days since S@H came back up (2/24AM in Berkeley), the Downloads have just been AWFUL. The current D/L constipation is completely painful - WUs now sit in Download:active status (usually in & out between active and Retry in 5,30,45 minutes or more) for interminably long time periods. Exiting and restarting BOINC previously helped get stalled WU downloads moving, but now it rarely gets any WU D/L to budge.

Maybe Matt L. needs to re-install his (unspecified) tweaks that got everything flowing so well 10 days ago? Please apply that much needed ExLax!
____________
Sabertooth Z77, i7-3770K@4.2GHz, GTX680, W8.1Pro x64
P5N32-E SLI, C2D E8400@3Ghz, GTX580, Win7SP1Pro x64 & PCLinuxOS2014 x64

TBar
Volunteer tester
Send message
Joined: 22 May 99
Posts: 1496
Credit: 53,066,946
RAC: 47,902
United States
Message 1340940 - Posted: 26 Feb 2013, 5:51:51 UTC

Maybe someone needs to end the ongoing Shortie Storm?
It's been going on for a while :^)

Let's see, one minute per file, 3 files per card, 4 cards per machine. It adds up...

Profile Donald L. JohnsonProject donor
Avatar
Send message
Joined: 5 Aug 02
Posts: 6324
Credit: 769,082
RAC: 942
United States
Message 1341026 - Posted: 26 Feb 2013, 13:44:49 UTC - in response to Message 1340940.

Maybe someone needs to end the ongoing Shortie Storm?
It's been going on for a while :^)

Let's see, one minute per file, 3 files per card, 4 cards per machine. It adds up...

May not be something they can control. Remember, Seti@Home gets its data by piggy-backing off other projects using the Arecibo telescope. We are at the mercy of the folks who pay to use the antenna - so we get what we get.

I don't know if the guys know what's on each set of "tapes" before they are fed to the splitters, but once they start splitting, they pretty much have to finish the run.....
____________
Donald
Infernal Optimist / Submariner, retired

Profile Tim
Volunteer tester
Avatar
Send message
Joined: 19 May 99
Posts: 205
Credit: 250,528,832
RAC: 54,206
Greece
Message 1341034 - Posted: 26 Feb 2013, 14:32:16 UTC - in response to Message 1340940.

Maybe someone needs to end the ongoing Shortie Storm?
It's been going on for a while :^)

Let's see, one minute per file, 3 files per card, 4 cards per machine. It adds up...


They can’t do that, but what they can is send sorties to CPU.

Tim

____________

N9JFE David SProject donor
Volunteer tester
Avatar
Send message
Joined: 4 Oct 99
Posts: 12450
Credit: 14,824,276
RAC: 4,619
United States
Message 1341035 - Posted: 26 Feb 2013, 14:32:19 UTC - in response to Message 1340826.

Question. Why did the server problem 81 get locked.

When a thread gets up to a few hundred posts, it gets locked and a new one is started, with the discussion continuing right where it left off. It is perfectly fine that you responded to something from the last one in this one.

____________
David
Sitting on my butt while others boldly go,
Waiting for a message from a small furry creature from Alpha Centauri.


Profile TRuEQ & TuVaLu
Volunteer tester
Avatar
Send message
Joined: 4 Oct 99
Posts: 479
Credit: 19,963,078
RAC: 19,958
Sweden
Message 1341068 - Posted: 26 Feb 2013, 16:48:41 UTC

dl 5 ap tasks for about 48hours now or something.
But they are almost complete. 80-98% now.

Rolf
Send message
Joined: 16 Jun 09
Posts: 114
Credit: 7,817,146
RAC: 0
Switzerland
Message 1341078 - Posted: 26 Feb 2013, 21:26:22 UTC - in response to Message 1341034.


They can’t do that, but what they can is send sorties to CPU.

Tim

That's what they do (for me) at the moment. But here is still the download speed (~2KBps)

Profile SonicAgamemnon
Avatar
Send message
Joined: 8 Apr 06
Posts: 31
Credit: 11,893,945
RAC: 0
United States
Message 1341117 - Posted: 27 Feb 2013, 0:14:56 UTC - in response to Message 1340826.

Question. Why did the server problem 81 get locked. The last post showed an awesome machine. So I will comment here. WOW, WOW,WOW, I want one. I was wondering how he had 32 threads going out a 8 core. Did not expect two cpus. It just looks so clean...


Here's a fun little video that shows the same machine munching on a CUDA shortie in just under 3.5 minutes. That is just enough time to hear my favorite guitar piece by Torroba during the video:

https://www.youtube.com/watch?v=OC9B3DsHY4w



I hope you enjoy watching this cute CUDA shortie climbing Mount SETI!
____________
"History is a pack of lies about events that never happened told by people who weren't there." - Santayana

Profile TRuEQ & TuVaLu
Volunteer tester
Avatar
Send message
Joined: 4 Oct 99
Posts: 479
Credit: 19,963,078
RAC: 19,958
Sweden
Message 1341160 - Posted: 27 Feb 2013, 5:15:40 UTC

5 more ap tasks in download.
Stalled ofcourse.

TBar
Volunteer tester
Send message
Joined: 22 May 99
Posts: 1496
Credit: 53,066,946
RAC: 47,902
United States
Message 1341165 - Posted: 27 Feb 2013, 5:43:11 UTC - in response to Message 1340781.
Last modified: 27 Feb 2013, 6:09:33 UTC

I gave up on connecting my XP machine to Ubuntu via Firewire. I finally broke down and installed Squid in Ubuntu. I actually managed to get Squid to handle the SETI downloads on my XP machine. No More Download Stalls with XP. I do need to look into configuring Squid to handle the SETI uploads and Scheduler requests though. Just like with many of the HTTP proxies, Squid doesn't do Uploads & Scheduler requests. Kinda makes you think many of those HTTP proxies are running Squid. My new HTTP Proxy is 192.168.1.4:5555.

Oh, no more of those 'Permanent HTTP Download Errors' either...

I've managed to find a solution to the Linux Proxies not working with the SETI Uploader & Scheduler. Just add this line to the "Don't use proxy for:" BOINC HTTP Proxy box;
http://setiboincdata.ssl.berkeley.edu/sah_cgi/file_upload_handler,http://setiboinc.ssl.berkeley.edu/sah_cgi/cgi

Then the Uploader & Scheduler will work when connected to a Proxy. All you have to do now is find a suitable Proxy. You could make your own, Mine downloads a Multibeam task in under a minute, an AstroPulse task in around 12-15 minutes...Without Any Stalls or Backoffs...

cdemers
Volunteer tester
Send message
Joined: 18 May 99
Posts: 29
Credit: 16,058,091
RAC: 724
Canada
Message 1341230 - Posted: 27 Feb 2013, 10:51:20 UTC

I just transparently proxied all port 80 traffic leaving my network using my router. Was sick and tired of playing with windows networking failing miserably with the latency problems. Have been downloading fine ever since and it has not been causing any scheduler contact problems. All tasks download at a resonable rate 10K-20K/s, which I am fine with, very rarely do I ever get a stalled download anymore.

This is my router BTW -> http://routerboard.com/RB751G-2HnD

____________

Wedge009
Volunteer tester
Avatar
Send message
Joined: 3 Apr 99
Posts: 356
Credit: 152,971,558
RAC: 81,290
Australia
Message 1341242 - Posted: 27 Feb 2013, 12:25:43 UTC - in response to Message 1341165.
Last modified: 27 Feb 2013, 12:26:01 UTC

I've managed to find a solution to the Linux Proxies not working with the SETI Uploader & Scheduler.

Thanks for that. After reading your post earlier today, I decided to give the local proxy a try, even though I couldn't think of any reason why it should work better than BOINC connecting to Berkeley directly. To my astonishment, the Squid proxy does seem to be a fair improvement - no more stalled downloads. Doesn't matter if the download rate is less than 10 KiB/s each, as long as it's making progress and I don't have to constantly manually kick it, this seems to be a good solution. It seems to me that Squid is a lot more robust in terms of retrieving data from Berkeley than from doing so with BOINC directly.
____________
Soli Deo Gloria

Richard HaselgroveProject donor
Volunteer tester
Send message
Joined: 4 Jul 99
Posts: 8759
Credit: 52,708,158
RAC: 26,743
United Kingdom
Message 1341251 - Posted: 27 Feb 2013, 12:51:09 UTC - in response to Message 1341242.

Mind you, my downloads today are running more smoothly than yesterday - and I haven't set up a proxy or changed any settings.

As has been frequently remarked, not having any AP tapes to split makes a big difference. Make sure you make the proper tie-up between cause and effect - and keep your proxy active during the next AP batch. That'll be the true test.

Wedge009
Volunteer tester
Avatar
Send message
Joined: 3 Apr 99
Posts: 356
Credit: 152,971,558
RAC: 81,290
Australia
Message 1341257 - Posted: 27 Feb 2013, 12:59:29 UTC

I did notice that downloads were marginally better in the last few hours. However, using my local proxy means that I'm getting about twice the speed that I was previously and I don't have to constantly hit the retry button all the time.

I'm aware that when AP WUs are being delivered, things slow to a crawl (hence my question about the compression in the other thread). However, I don't think things have stabilised to the point where all the AP WUs have all been downloaded yet - certainly I still have dozens of AP WUs to download, but at least they're making progress now.

At the very least, not having BOINC have such horrendously long back-offs (often longer than the time I'm at work) is a bonus, regardless of what the eventual download performance will be.
____________
Soli Deo Gloria

Previous · 1 · 2 · 3 · 4 · 5 . . . 24 · Next

Message boards : Number crunching : Panic Mode On (82) Server Problems?

Copyright © 2014 University of California