Communication constantly being deferred
15 Nov 2020 18:00:12 UTC
Topic 223986
(moderation:
One of my computers has had communication constantly deferred over the last 3 weeks or so, i have reset the project several time and uninstalled Boinc and reinstalled and this is still happening. Any help would be appreciated.
Cheers
Colin wrote: One of my
)
You didn't specify which of your 4 pc's has the problem but one thing I did notice is your that cache size seems to be too high as you have alot of 'not started by the deadline' tasks, reducing your cache size would help solve that problem. Also if you are running other tasks of the same type from other projects especially reduce it so both projects send you fewer tasks.
Thanks for the reply. It is
)
Thanks for the reply. It is the i7-4700 laptop, where would i reduce the cache size, wasn't able to find that anywhere.
Colin wrote:cache size If
)
If you were using web preferences then https://einsteinathome.org/account/prefs and computing ... advanced settings ... 'network usage' :
"Store at least:" and "Store up to an additional:"
OR
Boinc Manager ... Options ... Computing preferences...
'Computing' tab ... 'Other' ...
"Store at least:" and "Store up to an additional:"
Those are the places. That setting doesn't always instantly give a result that an alien mind would expect. Sometimes a very low setting might give no work at all and just a slightly higher setting could give a blast of work. But the threshold for "more work will be sent right now" is lingering somewhere there.
Problem solved, thanks all
)
Problem solved, thanks all for the help.
Cheers
Colin wrote:One of my
)
If you go to your account page on the website you will see your computers listed. Click the link for the one you are concerned about. Towards the bottom of the page that opens, you will see two clickable links for "Tasks" and "Last time contacted server" respectively. If you click the second one of those, you will see exactly why you are being "constantly deferred". The message to note is:-
[send] stopping work search - daily quota exceeded (24>=24)While you are on that page, also note the CPU work request was:-
[send] CPU: req 2419200.00 sec, 8.00 instances; est delay 0.00ie. a total of 28 days, which translates to 3.5 days for each of the 8 threads in the machine. There were separate large requests for CUDA (meaning your discrete GPU) and your internal Intel GPU. These large requests are not default values. At some point you must have set them previously. You really need to think carefully about changing default settings and at least how to reverse them if things get out of hand.
If you click the first of the two links I mentioned previously, you get to see all the tasks assigned to your computer. When I looked just now there were 200 that had all errored out in various ways and a brand new batch of 24 tasks for your discrete GPU. A 'new day' has obviously just ticked over for you so the scheduler has sent you your "daily allowance" (24 tasks) in one hit.
Unfortunately, they are all probably destined to fail (like all the previous ones) because your discrete GPU doesn't have sufficient memory. I can see this because I clicked the link for one of your previously failed tasks for the GW GPU search. I looked through the information returned to the project and found the following little snippet:-
Processing FFT failed: CL_MEM_OBJECT_ALLOCATION_FAILUREIn other words, when the GW GPU app tried to allocate sufficient memory for all the data in order to do the Fast Fourier Transform (FFT), it couldn't allocate sufficient memory.
As a general note, please understand that laptops are very likely to suffer premature failure if used to "crunch to the max" like this. In general, they are really not designed to cope with this huge level of stress. You would be very wise to change your preferences to limit how many 'compute devices' in the machine that BOINC is allowed to queue work for. You might be able to safely use your discrete GPU for the Gamma-ray Pulsar search (rather than the GW search), particularly if you disabled the Intel GPU and most (if not all) CPU threads.
If you would like to safely crunch with that machine, you need to spend some time understanding all the preferences and the various tools (just as I've used in the above discussion) to understand why things fail when they do. If you find things you don't understand, just ask, and someone is bound to point you in the right direction.
Cheers,
Gary.
Colin wrote:Problem solved,
)
Unfortunately, the problem isn't solved at all.
As I mentioned in my previous reply, the 24 new tasks have all failed with the same memory allocation error.
Please check for yourself.
EDIT: I just had a look via the "Last server contact" link for your computer and saw the return of the 24 newly failed tasks. You are still requesting the large amounts of work - denied because the daily quota is again exceeded. You won't be able to make another successful work request for a further 24 hours. This is likely to continue until you fix it properly.
Cheers,
Gary.
Not sure what has changed
)
Not sure what has changed with this setup, other than the amount of work requested that I changed on the 15th this computer has been running BOINC since I bought it in 2014 with no problems or changes. I put the work requested back to its original setting. I may try to unistall again and make sure I search for any files related to the program and pruge them to see if that helps.
Cheers
Colin wrote:Not sure what has
)
And what values do you now have? There are two settings, days and extra days. What do you have for each setting? Until the problem is resolved, I would suggest 0.2 days and 0.0 days respectively.
In 2014 and for many years after that, there were no Gravity Wave (GW) GPU tasks. There would have been no problem with the gamma-ray pulsar (GRP) tasks that existed then. You now have both these separate searches selected in your preferences and you urgently need to change your preferences to disable the GW GPU tasks. You are trashing huge numbers of them.
On the machine in question, you currently have 749 GW GPU tasks. There are already 491 that have errored out and 258 still in progress. These 258 will also error out as your machine tries to crunch them. You don't have a GPU with enough memory to be able to crunch them. There are zero of these GW GPU tasks that have succeeded on your machine. You need to stop requesting GW GPU work by changing your preferences to exclude that search. Uninstalling/reinstalling BOINC will have absolutely zero beneficial effect.
Cheers,
Gary.
Days is set to 2 and extra
)
Days is set to 2 and extra days is set to 4. I'll make the changes now.
Cheers
Just noticed on one of my
)
Just noticed on one of my other computers that it is only running one task instead of the 8 it normally runs. It's an i7-3770k not sure if its doing this for the reasons stated above as well. Any insights would be appreciated.
Cheers