Questions on Servers and WU's

Author	Message
Terror Australis Volunteer tester Send message Joined: 14 Feb 04 Posts: 1417 Credit: 181,743,642 RAC: 158,574	Message 913154 - Posted: 2 Jul 2009, 2:19:08 UTC
According to the Server status page as at 0200 UTC there is 1.4 million MB and nearly 5k AP WU's available for download. The Cricket graphs show the system is nowhere near max'ed out and the Server status page has a "green board". Q1)Why then do I keep getting "No jobs available" messages on my boxes ? Where and what is the problem ? Q2)One of my crunchers is out of CPU units but still has 60 odd CUDA units waiting to go, this is approx 36 hours of work. Is there any way to set the client for "crunch on first available processor", or any way to convert CUDA units to crunch on the CPU ? I realise these are old questions but I searched the fora and couldn't find the answers. TIA Brodo
ID: 913154 ·

Nobodi Send message Joined: 30 Nov 03 Posts: 13 Credit: 12,797,487 RAC: 386	Message 913160 - Posted: 2 Jul 2009, 3:06:02 UTC - in response to Message 913154.
I am seeing similar behaviour 1.4 million MB and nearly 5.7k AP WU's available for download. settings Run only the selected applications SETI@home Enhanced: no Astropulse: yes Astropulse v5: yes from other applications? yes When there was no ap work it was sending enhanced work at present it not sending any of that or the ap work.
ID: 913160 ·

1mp0£173 Volunteer tester Send message Joined: 3 Apr 99 Posts: 8423 Credit: 356,897 RAC: 0	Message 913162 - Posted: 2 Jul 2009, 3:20:04 UTC - in response to Message 913154. Last modified: 2 Jul 2009, 3:21:01 UTC
Q1)Why then do I keep getting "No jobs available" messages on my boxes ? Where and what is the problem ? The count is the total number of work units split, but not yet assigned. The splitters split the work. The feeder grabs 100 work units and those are available to the scheduler. If the feeder queue is empty (if the scheduler has assigned all of the 100 work unit batch) it will report "no work available" -- until the feeder grabs another 100 work units. Among other things, this keeps the download server from getting 100,000 requests to download all at once. ____________
ID: 913162 ·

Zeus Fab3r Send message Joined: 17 Jan 01 Posts: 639 Credit: 56,990,469 RAC: 81,245	Message 913171 - Posted: 2 Jul 2009, 3:51:48 UTC - in response to Message 913154.
Q2)One of my crunchers is out of CPU units but still has 60 odd CUDA units waiting to go, this is approx 36 hours of work. Is there any way to set the client for "crunch on first available processor", or any way to convert CUDA units to crunch on the CPU ? There is a thread at lunatics, about rebranding 6.03(CPU) into 6.08(GPU)WUs and vice versa. GPU rebranding There are couple solutions, but I'm using ReSchedule 1.7 cause it's standalone and easy to use. ____________ Who the hell is General Failure and why is he reading my harddisk?¿
ID: 913171 ·

Josef W. Segur Volunteer developer Volunteer tester Send message Joined: 30 Oct 99 Posts: 3932 Credit: 934,327 RAC: 15	Message 913187 - Posted: 2 Jul 2009, 4:35:44 UTC - in response to Message 913154.
According to the Server status page as at 0200 UTC there is 1.4 million MB and nearly 5k AP WU's available for download. The Cricket graphs show the system is nowhere near max'ed out and the Server status page has a "green board". Q1)Why then do I keep getting "No jobs available" messages on my boxes ? Where and what is the problem ? ... Brodo See Matt's post from last December for that issue. Just what is delaying the Feeder is a very good question, though. Joe
ID: 913187 ·

Terror Australis Volunteer tester Send message Joined: 14 Feb 04 Posts: 1417 Credit: 181,743,642 RAC: 158,574	Message 913195 - Posted: 2 Jul 2009, 5:42:51 UTC
All OK now with the rebranding. Thanks Zeus for the pointer to the rebranding program and to Ned and Joe (again) for the answers to the server question. Brodo
ID: 913195 ·

Terror Australis Volunteer tester Send message Joined: 14 Feb 04 Posts: 1417 Credit: 181,743,642 RAC: 158,574	Message 913235 - Posted: 2 Jul 2009, 10:21:39 UTC
That ReSchedule 1.7 looks like a nice little toy. I think I'm going to enjoy playing with that one :-) Thanks again Zeus and to the boys from Lunatics Brodo
ID: 913235 ·

Richard Haselgrove Volunteer tester Send message Joined: 4 Jul 99 Posts: 7458 Credit: 39,659,739 RAC: 33,292	Message 913276 - Posted: 2 Jul 2009, 14:28:28 UTC - in response to Message 913187.
See Matt's post from last December for that issue. Just what is delaying the Feeder is a very good question, though. Joe Something is slowing down the scheduler too. It doesn't usually take 86 seconds to get a null reponse: 02/07/2009 15:20:10\|SETI@home\|Sending scheduler request: To fetch work 02/07/2009 15:20:10\|SETI@home\|Requesting 368777 seconds of new work 02/07/2009 15:21:36\|SETI@home\|Scheduler RPC succeeded [server version 607] 02/07/2009 15:21:36\|SETI@home\|Message from server: (Project has no jobs available)
ID: 913276 ·

1mp0£173 Volunteer tester Send message Joined: 3 Apr 99 Posts: 8423 Credit: 356,897 RAC: 0	Message 913352 - Posted: 2 Jul 2009, 18:30:39 UTC - in response to Message 913276.
See Matt's post from last December for that issue. Just what is delaying the Feeder is a very good question, though. Joe Something is slowing down the scheduler too. It doesn't usually take 86 seconds to get a null reponse: 02/07/2009 15:20:10\|SETI@home\|Sending scheduler request: To fetch work 02/07/2009 15:20:10\|SETI@home\|Requesting 368777 seconds of new work 02/07/2009 15:21:36\|SETI@home\|Scheduler RPC succeeded [server version 607] 02/07/2009 15:21:36\|SETI@home\|Message from server: (Project has no jobs available) ... all the other traffic would do this. It could take 86 seconds just to get the SYN/SYN+ACK/ACK through a heavily loaded wire successfully. At times, the BOINC client is quite effectively doing a Distributed Denial of Service attack on the BOINC servers. ____________
ID: 913352 ·

Author

Message

Terror Australis
Volunteer tester
Send message
Joined: 14 Feb 04
Posts: 1417
Credit: 181,743,642
RAC: 158,574

According to the Server status page as at 0200 UTC there is 1.4 million MB and nearly 5k AP WU's available for download. The Cricket graphs show the system is nowhere near max'ed out and the Server status page has a "green board".

Q1)Why then do I keep getting "No jobs available" messages on my boxes ? Where and what is the problem ?

Q2)One of my crunchers is out of CPU units but still has 60 odd CUDA units waiting to go, this is approx 36 hours of work. Is there any way to set the client for "crunch on first available processor", or any way to convert CUDA units to crunch on the CPU ?

I realise these are old questions but I searched the fora and couldn't find the answers.

TIA
Brodo

ID: 913154 ·

Nobodi
Send message
Joined: 30 Nov 03
Posts: 13
Credit: 12,797,487
RAC: 386

I am seeing similar behaviour 1.4 million MB and nearly 5.7k AP WU's available for download.
settings
Run only the selected applications
SETI@home Enhanced: no
Astropulse: yes
Astropulse v5: yes
from other applications? yes

When there was no ap work it was sending enhanced work at present it not sending any of that or the ap work.

ID: 913160 ·

1mp0£173
Volunteer tester
Send message
Joined: 3 Apr 99
Posts: 8423
Credit: 356,897
RAC: 0

Q1)Why then do I keep getting "No jobs available" messages on my boxes ? Where and what is the problem ?

The count is the total number of work units split, but not yet assigned.

The splitters split the work.

The feeder grabs 100 work units and those are available to the scheduler.

If the feeder queue is empty (if the scheduler has assigned all of the 100 work unit batch) it will report "no work available" -- until the feeder grabs another 100 work units.

Among other things, this keeps the download server from getting 100,000 requests to download all at once.
____________

ID: 913162 ·

Zeus Fab3r

Send message
Joined: 17 Jan 01
Posts: 639
Credit: 56,990,469
RAC: 81,245

Q2)One of my crunchers is out of CPU units but still has 60 odd CUDA units waiting to go, this is approx 36 hours of work. Is there any way to set the client for "crunch on first available processor", or any way to convert CUDA units to crunch on the CPU ?

There is a thread at lunatics, about rebranding 6.03(CPU) into 6.08(GPU)WUs and vice versa.

GPU rebranding

There are couple solutions, but I'm using ReSchedule 1.7 cause it's standalone and easy to use.
____________

Who the hell is General Failure and why is he reading my harddisk?¿

ID: 913171 ·

Josef W. Segur

Volunteer developer
Volunteer tester
Send message
Joined: 30 Oct 99
Posts: 3932
Credit: 934,327
RAC: 15

According to the Server status page as at 0200 UTC there is 1.4 million MB and nearly 5k AP WU's available for download. The Cricket graphs show the system is nowhere near max'ed out and the Server status page has a "green board".

Q1)Why then do I keep getting "No jobs available" messages on my boxes ? Where and what is the problem ?
...
Brodo

See Matt's post from last December for that issue. Just what is delaying the Feeder is a very good question, though.

Joe

ID: 913187 ·

Terror Australis
Volunteer tester
Send message
Joined: 14 Feb 04
Posts: 1417
Credit: 181,743,642
RAC: 158,574

All OK now with the rebranding. Thanks Zeus for the pointer to the rebranding program and to Ned and Joe (again) for the answers to the server question.

Brodo

ID: 913195 ·

Terror Australis
Volunteer tester
Send message
Joined: 14 Feb 04
Posts: 1417
Credit: 181,743,642
RAC: 158,574

That ReSchedule 1.7 looks like a nice little toy. I think I'm going to enjoy playing with that one :-) Thanks again Zeus and to the boys from Lunatics

Brodo

ID: 913235 ·

Richard Haselgrove

Volunteer tester
Send message
Joined: 4 Jul 99
Posts: 7458
Credit: 39,659,739
RAC: 33,292

See Matt's post from last December for that issue. Just what is delaying the Feeder is a very good question, though.
Joe

Something is slowing down the scheduler too. It doesn't usually take 86 seconds to get a null reponse:

ID: 913276 ·

1mp0£173
Volunteer tester
Send message
Joined: 3 Apr 99
Posts: 8423
Credit: 356,897
RAC: 0

See Matt's post from last December for that issue. Just what is delaying the Feeder is a very good question, though.
Joe

Something is slowing down the scheduler too. It doesn't usually take 86 seconds to get a null reponse:

02/07/2009 15:20:10|SETI@home|Sending scheduler request: To fetch work
02/07/2009 15:20:10|SETI@home|Requesting 368777 seconds of new work
02/07/2009 15:21:36|SETI@home|Scheduler RPC succeeded [server version 607]
02/07/2009 15:21:36|SETI@home|Message from server: (Project has no jobs available)

... all the other traffic would do this. It could take 86 seconds just to get the SYN/SYN+ACK/ACK through a heavily loaded wire successfully.

At times, the BOINC client is quite effectively doing a Distributed Denial of Service attack on the BOINC servers.
____________

ID: 913352 ·