Thanks for the input. I have lowered the cache size as suggested and will monitor the GPU concurrent job count. Interesting info on the NVIDIA side Gary.
Changing the cache size has restored normal operation. I remember changing to "2" days back when E@H experienced that one week outage. I panicked and made the change w/o understanding its full effect. Now I have undone that change on all machines. In the time I have been with E@H I have only experienced that one outage and I panicked. Shame on me.
It seems to be prioritizing a bit too much, as it stops running one of the 3 gpu concurrent tasks. So I only have 2 GPU task running at the moment even though there is meant to be 3.
I can only conclude that as I have AMD GPU which normally use .5 of the CPU for each task, it has decided to do a CPU job instead of a GPU job. Didn't know it could do that.
The exact same thing happened to me on NVIDIA cards. Both cards went from 3 at a time to 2 at a time.
This is NOT a complaint. I'm fine with what happened. My cache size was 1 day and I received a download of enough GW follow-ups to take two days (plus a little more).
You might ask, "What did you do wrong?" Well, what I did was tell BOINC it could only use 50% of the processors. I did that because the AMD FX CPU only has half as many FPUs as "cores," so I wanted an FPU available per CPU work unit.
The machine decided that I needed to drop my _G_PU concurrent work from 6 to 4 almost as though it "knew" there weren't enough resources for all of those tasks or as though it assigned more than one CPU core to each task.
WHICH IS FINE! I am NOT complaining. I thought BOINC had become very, very clever.
Without touching a thing in the client, I discontinued allowing CPU tasks in that "Location." As soon as the CPU tasks finished the machine began doing 6 GPU tasks again.
I am NOT complaining. That behavior was fine by me. I wanted to do some of the CPU follow-up tasks to help clear them quickly. I don't think this needs to be "fixed."
I'm just reporting that it appeared (even if by coincidence)someone had been doing something server-side that was ingenious. I see others noticed the same thing.
I don't get any new GW tasks either so there is nothing wrong on your end. The server seems to be out of tasks to send.
It's a bit weird since the server status page says 15% remaining on the run.
Have you ticked the box for GW #3 in your Einstein@home preferences.
By default I don't think you will get any task in a new run until you change your preferences.
Changing the cache size has restored normal operation. I remember changing to "2" days back when E@H experienced that one week outage. I panicked and made the change w/o understanding its full effect. Now I have undone that change on all machines. In the time I have been with E@H I have only experienced that one outage and I panicked. Shame on me.
I wouldn't consider you panicked at all :-). In the normal course of things a 2-3 day cache is a fairly minimalist setting. The real problem was the sudden introduction of very short deadline work without some serverside mechanism to drip feed the tasks. The only way for the user to cope is to temporarily 'turn off the tap' through cache settings, possibly combined with temporary suspension of other science runs if the aim is to maximise participation in FU3. Now that FU3 seems to be largely done, don't forget to put your cache back up to where it was. Friday night problems that don't get fixed until Monday can happen :-).
The exact same thing happened to me on NVIDIA cards. Both cards went from 3 at a time to 2 at a time.
I don't have any machine with more than 1 GPU so I'm presuming rather than knowing for sure. With 6 GPU tasks running, that's 1.2 CPU cores required for support so with the 50% allowed CPU cores setting, I presume you would have 3 CPU tasks running normally with the 6 GPU tasks. Is that correct?
If panic mode is entered, an extra CPU task would be started and this would cause the allowed GPU tasks to drop to 4 (since 5 still uses up exactly 1 core). Because you already have 4 cores that BOINC isn't allowed to use, you have plenty of GPU support so you could have kept 6 GPU tasks running by using app_config.xml to change the default cpu_usage from 0.2 to say 0.15. With 6x0.15=0.9, panic mode would not be able to get another available CPU core by limiting the number of GPU tasks to 4.
Hallo!
Meanwhile I got 3 tasks and crunched them instantly and successfully. The last I reported yesterday 16:34 UTC. Since than there was more than 37 inquieries for new work that became answered with: No work available for this project. But in the Server Status you can see an increasing number of tasks to send oi more than 500. This number is increasing by 10 to 15 within 5 minutes. That mean to mee, that there became mor tasks available than are downloaded, but why not to me ???????
Why don´t I get new tasks, even when they are obviously available. ?????
RE: Gary/Richard, Thanks
)
Changing the cache size has restored normal operation. I remember changing to "2" days back when E@H experienced that one week outage. I panicked and made the change w/o understanding its full effect. Now I have undone that change on all machines. In the time I have been with E@H I have only experienced that one outage and I panicked. Shame on me.
RE: It seems to be
)
The exact same thing happened to me on NVIDIA cards. Both cards went from 3 at a time to 2 at a time.
This is NOT a complaint. I'm fine with what happened. My cache size was 1 day and I received a download of enough GW follow-ups to take two days (plus a little more).
You might ask, "What did you do wrong?" Well, what I did was tell BOINC it could only use 50% of the processors. I did that because the AMD FX CPU only has half as many FPUs as "cores," so I wanted an FPU available per CPU work unit.
The machine decided that I needed to drop my _G_PU concurrent work from 6 to 4 almost as though it "knew" there weren't enough resources for all of those tasks or as though it assigned more than one CPU core to each task.
WHICH IS FINE! I am NOT complaining. I thought BOINC had become very, very clever.
Without touching a thing in the client, I discontinued allowing CPU tasks in that "Location." As soon as the CPU tasks finished the machine began doing 6 GPU tasks again.
I am NOT complaining. That behavior was fine by me. I wanted to do some of the CPU follow-up tasks to help clear them quickly. I don't think this needs to be "fixed."
I'm just reporting that it appeared (even if by coincidence)someone had been doing something server-side that was ingenious. I see others noticed the same thing.
Hallo! Why don´t I get any
)
Hallo!
Why don´t I get any tasks from S6BucketFU3UB ??? What´s wrong in my adjustments?
Kind regards and happy crunching
Martin
I don't get any new GW tasks
)
I don't get any new GW tasks either so there is nothing wrong on your end. The server seems to be out of tasks to send.
It's a bit weird since the server status page says 15% remaining on the run.
Have you ticked the box for
)
Have you ticked the box for GW #3 in your Einstein@home preferences.
By default I don't think you will get any task in a new run until you change your preferences.
John
@astro Read the science
)
@astro
Read the science forum.
plans for near future E@h
There will be no more work for GW Search until data for the new advanced detectors arrived later this year.
In the mean time, there is a lot of FGRP 4 search to do.
Filipe
RE: Hallo! Why don´t I get
)
I think the answer is the same as mine...
Tue 15 Sep 2015 21:36:59 BST | Einstein@Home | No work is available for Gravitational Wave search S6Bucket Follow-up #3
Also http://einstein.phys.uwm.edu/server_status.html shows nearly done.
RE: Changing the cache size
)
I wouldn't consider you panicked at all :-). In the normal course of things a 2-3 day cache is a fairly minimalist setting. The real problem was the sudden introduction of very short deadline work without some serverside mechanism to drip feed the tasks. The only way for the user to cope is to temporarily 'turn off the tap' through cache settings, possibly combined with temporary suspension of other science runs if the aim is to maximise participation in FU3. Now that FU3 seems to be largely done, don't forget to put your cache back up to where it was. Friday night problems that don't get fixed until Monday can happen :-).
Cheers,
Gary.
RE: The exact same thing
)
I don't have any machine with more than 1 GPU so I'm presuming rather than knowing for sure. With 6 GPU tasks running, that's 1.2 CPU cores required for support so with the 50% allowed CPU cores setting, I presume you would have 3 CPU tasks running normally with the 6 GPU tasks. Is that correct?
If panic mode is entered, an extra CPU task would be started and this would cause the allowed GPU tasks to drop to 4 (since 5 still uses up exactly 1 core). Because you already have 4 cores that BOINC isn't allowed to use, you have plenty of GPU support so you could have kept 6 GPU tasks running by using app_config.xml to change the default cpu_usage from 0.2 to say 0.15. With 6x0.15=0.9, panic mode would not be able to get another available CPU core by limiting the number of GPU tasks to 4.
Cheers,
Gary.
Hallo! Meanwhile I got 3
)
Hallo!
Meanwhile I got 3 tasks and crunched them instantly and successfully. The last I reported yesterday 16:34 UTC. Since than there was more than 37 inquieries for new work that became answered with: No work available for this project. But in the Server Status you can see an increasing number of tasks to send oi more than 500. This number is increasing by 10 to 15 within 5 minutes. That mean to mee, that there became mor tasks available than are downloaded, but why not to me ???????
Why don´t I get new tasks, even when they are obviously available. ?????
Kind regards and happy crunching
Martin