Server Status
log in

Advanced search

Message boards : Number crunching : Server Status

Previous · 1 · 2 · 3 · 4 · 5 · 6 · 7 . . . 18 · Next
Author Message
Profile STE\/E
Avatar
Send message
Joined: 12 Jul 09
Posts: 581
Credit: 761,710,729
RAC: 0
Message 1916 - Posted: 24 Sep 2009, 9:25:52 UTC

I think it's Validator Kicking time ... :)

Profile mikey
Avatar
Send message
Joined: 11 Aug 09
Posts: 3241
Credit: 1,683,428,401
RAC: 7,688,895
Message 1918 - Posted: 24 Sep 2009, 11:23:18 UTC - in response to Message 1908.

Which manual should I be reading?

----- Rom


Hopefully the one you wrote when you programmed Boinc. You did write one...right?! I hope they gave you enough time to write one with all the work you did on Boinc!!!!

Profile Slicker
Volunteer moderator
Project administrator
Project developer
Project tester
Project scientist
Avatar
Send message
Joined: 11 Jun 09
Posts: 2525
Credit: 740,580,099
RAC: 2
Message 1926 - Posted: 24 Sep 2009, 14:38:48 UTC - in response to Message 1906.

OK -- the outage was short -- and the home page plus the threads here alerted folks to the possibility as well.

Regarding RTFM to do things right -- pass that on to the folks writing the BOINC client <smile>.


Sorry I didn't give more advance warning about the server being down, but when I got home this evening the server had just about enough disk space left to log the shutdown process.

So, I took the plunge and added the drive this evening and migrated the database to the new drive. All is well for now. This weekend I plan to change the database over to use one innodb file per table so when a table is optimized it will actually release space. The WU and result tables were the culprits as they get fragmented very easily. Guess I need to RTFM before assuming the default settings during an install are the best ones to use.



I was actually referring to the section in the mysql manual about managing and optimizing innodb files and the various config options. Or, had I read the BOINC wiki page that explained how to deal with DLLs and other files with the same name, I could have avoided most of the failed WUs that were due to the clients not downloading the new cuda run time DLL. Either probably would have kept the database size at a reasonable level. Both were documented, I just hadn't read either.

John Clark
Send message
Joined: 21 Sep 09
Posts: 548
Credit: 56,516,565
RAC: 0
Message 1946 - Posted: 25 Sep 2009, 15:37:49 UTC

Slicker

I presume your warning of the server going down this week end is still relevant, or has the disk fix sorted this requirement?

Profile Slicker
Volunteer moderator
Project administrator
Project developer
Project tester
Project scientist
Avatar
Send message
Joined: 11 Jun 09
Posts: 2525
Credit: 740,580,099
RAC: 2
Message 1947 - Posted: 25 Sep 2009, 16:06:50 UTC - in response to Message 1946.

Slicker

I presume your warning of the server going down this week end is still relevant, or has the disk fix sorted this requirement?


I still plan to take it down for a couple hours this weekend to change it so the innodb tables each have their own file rather than one for all tables. I believe that will give the best performance while minimizing the disk space used.

Profile BarryAZ
Send message
Joined: 21 Aug 09
Posts: 251
Credit: 13,167,700,808
RAC: 23,581,298
Message 1949 - Posted: 25 Sep 2009, 16:55:45 UTC - in response to Message 1908.

The one that talks about taking advantage of the project and account definition information and thus performing a targeted work fetch --- that is, to fetch ATI GPU work when the workstation/account is configured for that, to fetch GPU work with the workstation/account is configured for that, to fetch CPU only work when the workstation/account is configured for that, and to configure some mix only when the workstation/account is configured for that.

With the newest iterations of the BOINC client, the work fetch is generating excess traffic because it tries to get GPU work from CPU only projects as well as CPU work from a workstation/account configured as GPU work only. It even pings CUDA only projects for ATI GPU work. I've been informed this is a works the right way as designed -- I disagree with that view.


Which manual should I be reading?

----- Rom


OK -- the outage was short -- and the home page plus the threads here alerted folks to the possibility as well.

Regarding RTFM to do things right -- pass that on to the folks writing the BOINC client <smile>.


Rom Walton (BOINC)
Send message
Joined: 22 Sep 09
Posts: 81
Credit: 3,184,321
RAC: 0
Message 1953 - Posted: 25 Sep 2009, 19:36:48 UTC - in response to Message 1949.

Folowing up here:
http://boinc.thesonntags.com/collatz/forum_thread.php?id=113

----- Rom

The one that talks about taking advantage of the project and account definition information and thus performing a targeted work fetch --- that is, to fetch ATI GPU work when the workstation/account is configured for that, to fetch GPU work with the workstation/account is configured for that, to fetch CPU only work when the workstation/account is configured for that, and to configure some mix only when the workstation/account is configured for that.

With the newest iterations of the BOINC client, the work fetch is generating excess traffic because it tries to get GPU work from CPU only projects as well as CPU work from a workstation/account configured as GPU work only. It even pings CUDA only projects for ATI GPU work. I've been informed this is a works the right way as designed -- I disagree with that view.


Which manual should I be reading?

----- Rom


OK -- the outage was short -- and the home page plus the threads here alerted folks to the possibility as well.

Regarding RTFM to do things right -- pass that on to the folks writing the BOINC client <smile>.



Profile Chris S
Avatar
Send message
Joined: 12 Jul 09
Posts: 257
Credit: 89,016,367
RAC: 0
Message 1957 - Posted: 25 Sep 2009, 19:59:45 UTC

I still plan to take it down for a couple hours this weekend to change it so the innodb tables each have their own file rather than one for all tables. I believe that will give the best performance while minimizing the disk space used.


Thanks for the heads-up there. :-)

Profile Slicker
Volunteer moderator
Project administrator
Project developer
Project tester
Project scientist
Avatar
Send message
Joined: 11 Jun 09
Posts: 2525
Credit: 740,580,099
RAC: 2
Message 2006 - Posted: 26 Sep 2009, 20:38:59 UTC

For those who care....
I converted all the innodb tables to myisam. That worked fine. I then switched the server to use one innodb file per table and restarted it. So far, so good. Then I converted each table from myisam back to innodb. Again, all is well. Then I stopped the database, deleted the old, what I thought was not longer used, data file and restarted the database. It proceeded to create a new, small file. Just what I wanted. Only problem was, it decided the innodb tables I had just converted didn't exist any more. (I guess I should have deleted the file before switching back from myISAM, but since the network was down, I couldn't google for help.) Worse yet, it wouldn't let me drop them. It didn't want to drop the database either since the tables were there but not really there. So, forced it to drop the database, recreated it, and then loaded all the data using one of the two full backups I did before I started. Anyway, all is well again.

Profile STE\/E
Avatar
Send message
Joined: 12 Jul 09
Posts: 581
Credit: 761,710,729
RAC: 0
Message 2007 - Posted: 26 Sep 2009, 20:41:30 UTC

Good Job, didn't even really notice the Site was down very long ...

Profile Paul D. Buck
Volunteer tester
Send message
Joined: 30 Aug 09
Posts: 412
Credit: 185,735,226
RAC: 0
Message 2008 - Posted: 26 Sep 2009, 20:43:05 UTC - in response to Message 2006.

For those who care....
I converted all the innodb tables to myisam. That worked fine. I then switched the server to use one innodb file per table and restarted it. So far, so good. Then I converted each table from myisam back to innodb. Again, all is well. Then I stopped the database, deleted the old, what I thought was not longer used, data file and restarted the database. It proceeded to create a new, small file. Just what I wanted. Only problem was, it decided the innodb tables I had just converted didn't exist any more. (I guess I should have deleted the file before switching back from myISAM, but since the network was down, I couldn't google for help.) Worse yet, it wouldn't let me drop them. It didn't want to drop the database either since the tables were there but not really there. So, forced it to drop the database, recreated it, and then loaded all the data using one of the two full backups I did before I started. Anyway, all is well again.

That is why I usually MOVE the files so it can't find them before I delete them ...

I am just glad the back up worked. My experience with back-ups, even ones that validated is less than stellar ... though I have to admit that Oracle exports almost always worked ...

The good news is that I had worked off a bunch of the tasks I had in hand while MW and you were down ... so in a day or two I may be able to try to see if I can get BOINC to play nice with two GPU projects.

Profile Logan
Avatar
Send message
Joined: 2 Jul 09
Posts: 124
Credit: 37,455,338
RAC: 0
Message 2678 - Posted: 9 Oct 2009, 17:52:19 UTC

Hi Slicker, my friend!!!

I think the validator has a lot of homeworks (is down).

I don't know if you're updating it or not, so I must to warn you about that...;)
____________
Logan.

BOINC FAQ Service (Ahora, también disponible en Español/Now available in Spanish)

frankhagen
Send message
Joined: 12 Jul 09
Posts: 188
Credit: 14,210,374
RAC: 1,303
Message 2679 - Posted: 9 Oct 2009, 17:54:24 UTC - in response to Message 2678.

Hi Slicker, my friend!!!

I think the validator has a lot of homeworks (is down).


probably his boots are worn out now having to kick it so often.. ;)

____________

Profile BarryAZ
Send message
Joined: 21 Aug 09
Posts: 251
Credit: 13,167,700,808
RAC: 23,581,298
Message 2687 - Posted: 9 Oct 2009, 18:52:33 UTC - in response to Message 2678.

Yes it does seem that way from here -- updating doesn't seem to be going on at the moment.

Hi Slicker, my friend!!!

I think the validator has a lot of homeworks (is down).

I don't know if you're updating it or not, so I must to warn you about that...;)

Profile Slicker
Volunteer moderator
Project administrator
Project developer
Project tester
Project scientist
Avatar
Send message
Joined: 11 Jun 09
Posts: 2525
Credit: 740,580,099
RAC: 2
Message 2689 - Posted: 9 Oct 2009, 19:19:34 UTC

Some of the really old WUs that caused validator problems were resent and are now causing another round of validator issues now that the 2.0x wingman is returning the other half. If the validator didn't like it the first time around, it doesn't like it the next time around either.

Anyway, I gave it a kick. Let's hope it lasts. I'm on the hunt for bambi's older brother again this weekend.

Profile Logan
Avatar
Send message
Joined: 2 Jul 09
Posts: 124
Credit: 37,455,338
RAC: 0
Message 2690 - Posted: 9 Oct 2009, 19:25:33 UTC - in response to Message 2689.

Some of the really old WUs that caused validator problems were resent and are now causing another round of validator issues now that the 2.0x wingman is returning the other half. If the validator didn't like it the first time around, it doesn't like it the next time around either.

Anyway, I gave it a kick. Let's hope it lasts. I'm on the hunt for bambi's older brother again this weekend.


Thanks Slicker!!!

Give to Bambi's brother a gun to have one opportunity...:D
____________
Logan.

BOINC FAQ Service (Ahora, también disponible en Español/Now available in Spanish)

Odd-Rod
Send message
Joined: 11 Jul 09
Posts: 528
Credit: 3,118,537
RAC: 324
Message 2702 - Posted: 10 Oct 2009, 5:47:21 UTC - in response to Message 2679.

Hi Slicker, my friend!!!

I think the validator has a lot of homeworks (is down).


probably his boots are worn out now having to kick it so often.. ;)


If we can't donate new computer equipment, perhaps we could have a fund-raiser for a new pair of boots?
:D

Profile robertmiles
Send message
Joined: 8 Oct 09
Posts: 45
Credit: 14,857,029
RAC: 45,491
Message 2703 - Posted: 10 Oct 2009, 6:05:32 UTC

Slicker,

Some of the messages in this thread look like you need to talk to Comcast to make sure they have given your server a static IP address instead of a dynamic IP address. If it's a dynamic IP address, they're allowed to change the IP address just about any time you don't have a connection in progress, although they're likely to have some policy on just how often they actually change it.

Connections with static IP addresses often cost more than similar connections with dynamic IP addresses, though.

Profile Bruce
Avatar
Send message
Joined: 15 Jul 09
Posts: 294
Credit: 12,203,827
RAC: 0
Message 2728 - Posted: 11 Oct 2009, 13:27:52 UTC

Validator is down again.
____________

Profile Slicker
Volunteer moderator
Project administrator
Project developer
Project tester
Project scientist
Avatar
Send message
Joined: 11 Jun 09
Posts: 2525
Credit: 740,580,099
RAC: 2
Message 2773 - Posted: 12 Oct 2009, 14:19:10 UTC

The validator will continue fail until we a new set of ATI apps are released. Part of the 2.xx app release was to change to a different type of encoding. For whatever reason, boinc sometimes thinks it didn't upload the result file and sends some/all of the data a second time. The only problem is that some of the data did arrive with the first upload and rather than overwriting what is there, it appends the second copy to the original file. So, now the server has a corrupted result file with somewhere between 1 and 2 times the data it should have. When that gets run through the validator, it crashes not only the WU but the daemon fails which requires me to manually grant credit, then cancel the WU, then restart the validator.

Previous · 1 · 2 · 3 · 4 · 5 · 6 · 7 . . . 18 · Next
Post to thread

Message boards : Number crunching : Server Status


Main page · Your account · Message boards


Copyright © 2018 Jon Sonntag; All rights reserved.