Pending Task Count at 252 - Why So High?
14 Oct 2023 8:02:40 UTC
Topic 230200
(moderation:
I've been wondering why my daily credits have been lower for a few weeks. Checking my tasks, there are 252 pending validations, some as old as September 20. The oldest ones are Binary Radio Pulsar Search and Gamma-ray pulsar search #5. Of the ones I've glanced most are either from my MacBook Pro M1 (arm64-apple-darwin) or from my Jetson Orin.
Why are so many pending?
Is there a deadline for pending tasks where my crunching will be thrown out?
Besides "patience, young grasshopper" is there anything I can do to trigger validations?
TIA,
Sam
CyborgSam wrote:I've been
)
Hi CyborgSam,
To answer your question of "is there a deadline for pending tasks", no, there is not. The 'pending' means that YOU have completed the task but you are still waiting for your wingman to complete his duplication of the same task.
This is an unfortunate way (for you) to do things, i.e. "patience, young grasshopper", while at the same time it is very fortunate for Einstein@Home to validate the information in the task.
To answer your next question of "is there anything I can do to trigger validations?" the answer is no, there isn't.
I hope this helps.
.....[EDIT].....
Just as a reference, my two computers in use at the moment have 282 & 152 pending tasks. No big deal!
Proud member of the Old Farts Association
George-> Thanks. My main
)
George->
Thanks. My main concern is the age of the pending tasks, going back to September 20th. I assume (with no knowledge) that a wingman must be capable of running the exact same app I did. So could it be that there aren't enough Mac M1 and Jetson Orin users to handle the validations?
And why would the amount of pending tasks increase so much in the last three weeks?
TIA, Sam
There is no scheduler
)
There is no scheduler mechanism that matches host with like host. Your wingmen is from the luck of the draw of whomever is up next.
The inconclusives and pendings are increasing because of the dropping of the FGRPB1G tasks and apps and the subsequent significant loss of crunching power from crunchers leaving the project.
That's too bad crunchers are
)
That's too bad crunchers are leaving the project. Is it because they're getting lower credits without FGRPB1G?
Hard to say but likely the
)
Hard to say but likely the most popular reason. The only other gpu apps are very slow and inefficient compared to the old GR#1 tasks.
I don't think they all left.
)
I don't think they all left. I had a pending of over 4000 with the AS Gravity tasks for a bit but when the gamma ray app ran out it dropped rapidly. Could just a coincidence though.
After patiently waiting for a
)
After patiently waiting for a week, the tasks I originally posted about are still pending. My Macs reported them on September 21, over a month ago. There are other tasks over two weeks old.
There are plenty of newer valid tasks listed.
Looking at the Work Unit ID, I don't see a wingman listed anywhere. Does this get added after the task is no longer pending?
TIA,
Sam
I kinda have a suspcion that
)
I kinda have a suspcion that no outstanding wingmen tasks are being issued to pending work. Simply because there IS NO work anymore.
My Mac Pro w/Xeons is doing
)
My Mac Pro w/Xeons is doing BRP tasks and my M1 MacBook Pro is doing GRP tasks. The oldest tasks that haven't been validated were run on my M1 MacBook Pro and my Jetson Orin. Were these the "old" BRPs?
My new MacBook Air 15" only got 2200 credits yesterday after running just over 24 hours. It'll take a few days to see how it's performing overall since the Einstein and PrimeGrid tasks take so long. It's staying cool enough though, I have both MacBooks on a laptop cooler with fans.
CyborgSam wrote:.... The
)
Have you tried, using the TaskID links on the website, to look at what was sent back to the project for any of the outstanding tasks from that machine?
I just looked through the stderr output that was logged for this task. If you just read the log you can see that there were a total of 8 attempts for
which were followed by lines of header information and within a minute or two
This seems rather fast but the 8 cycles are to be expected since these 'large' tasks are a combination of smaller tasks (8 if my memory is correct). What is not expected is that no checkpoints were ever created, there were no indications of any errors occurring, and there was no indication of any output results being produced. The log simply shows the transition to the next small task in the list of 8 with the very next line saying
Each of the 8 small tasks should show the creation of checkpoints during the computation and at the end of each cycle you would expect to see some sort of results being created. I don't run these tasks so I don't have anything to compare with to know for sure.
After the 8 small tasks are done, you would expect to see some sort of results file(s) being created and uploaded, or at least some sort of error message. Unfortunately, the lack of such files seems to mean there never were any returned results and those tasks can't ever be validated. If you are unhappy about that, you will need to message Bernd about it.
It's impossible to see the workunit info for these tasks. The contents are always hidden until the workunit is validated. It's quite possible that these workunits were a 'quorum of 1' where the results of 'trusted' hosts could be validated without needing to be checked against a second set of results. Once again, this is just speculation about what might be happening and Bernd is the only person who can say for sure.
Cheers,
Gary.