No new Work units for over a week
19 Jan 2016 6:47:20 UTC
Topic 198385
(moderation:
So I no longer seem to be getting work units for my iMac as of last week. I can't tell if it is a Work Generator issue (Einstein out of work) or if there is a issue with my computer. I am posting here because some of the messages in the last communication seem worrisome as they talk of "trust"issues. I've been crunching for 10 years and never had a problem. I also see a message about Aericbo not available which is fine. Last units I was working were Gravitational waves.
Thanks
Here is the message traffic:
2016-01-19 06:31:06.0126 [PID=12895] Request: [USER#xxxxx] [HOST#2242651] [IP xxx.xxx.xxx.80] client 7.6.22
2016-01-19 06:31:06.0818 [PID=12895] [send] effective_ncpus 8 max_jobs_on_host_cpu 999999 max_jobs_on_host 999999
2016-01-19 06:31:06.0818 [PID=12895] [send] effective_ngpus 1 max_jobs_on_host_gpu 999999
2016-01-19 06:31:06.0818 [PID=12895] [send] Not using matchmaker scheduling; Not using EDF sim
2016-01-19 06:31:06.0818 [PID=12895] [send] CPU: req 0.00 sec, 0.00 instances; est delay 0.00
2016-01-19 06:31:06.0818 [PID=12895] [send] ATI: req 86400.00 sec, 1.00 instances; est delay 0.00
2016-01-19 06:31:06.0818 [PID=12895] [send] work_req_seconds: 0.00 secs
2016-01-19 06:31:06.0818 [PID=12895] [send] available disk 99.70 GB, work_buf_min 43200
2016-01-19 06:31:06.0819 [PID=12895] [send] active_frac 0.999972 on_frac 0.759640 DCF 1.000000
2016-01-19 06:31:06.0827 [PID=12895] [send] [HOST#2242651] not reliable; max_result_day 31
2016-01-19 06:31:06.0828 [PID=12895] [send] set_trust: error rate 0.090259 > 0.050000, don't trust
2016-01-19 06:31:06.0828 [PID=12895] [mixed] sending non-locality work first (0.9740)
2016-01-19 06:31:06.1004 [PID=12895] [version] Checking plan class 'FGRP4-Beta'
2016-01-19 06:31:06.1026 [PID=12895] [version] reading plan classes from file '/BOINC/projects/EinsteinAtHome/plan_class_spec.xml'
2016-01-19 06:31:06.1026 [PID=12895] [version] beta test app versions not allowed in project prefs.
2016-01-19 06:31:06.1027 [PID=12895] [version] Checking plan class 'FGRP4-SSE2'
2016-01-19 06:31:06.1027 [PID=12895] [version] plan class ok
2016-01-19 06:31:06.1027 [PID=12895] [version] Don't need CPU jobs, skipping version 115 for hsgamma_FGRP4 (FGRP4-SSE2)
2016-01-19 06:31:06.1027 [PID=12895] [version] no app version available: APP#27 (hsgamma_FGRP4) PLATFORM#10 (x86_64-apple-darwin) min_version 0
2016-01-19 06:31:06.1027 [PID=12895] [version] no app version available: APP#27 (hsgamma_FGRP4) PLATFORM#6 (i686-apple-darwin) min_version 0
2016-01-19 06:31:06.1028 [PID=12895] [version] no app version available: APP#19 (einsteinbinary_BRP4) PLATFORM#10 (x86_64-apple-darwin) min_version 0
2016-01-19 06:31:06.1029 [PID=12895] [version] no app version available: APP#19 (einsteinbinary_BRP4) PLATFORM#6 (i686-apple-darwin) min_version 0
2016-01-19 06:31:06.1036 [PID=12895] [version] Checking plan class 'BRP6-cuda55-Lion'
2016-01-19 06:31:06.1037 [PID=12895] [version] parsed project prefs setting 'gpu_util_brp': 0.000000
2016-01-19 06:31:06.1037 [PID=12895] [version] No CUDA devices found
2016-01-19 06:31:06.1037 [PID=12895] [version] Checking plan class 'BRP6-Beta-opencl-ati-lion'
2016-01-19 06:31:06.1037 [PID=12895] [version] beta test app versions not allowed in project prefs.
2016-01-19 06:31:06.1037 [PID=12895] [version] Checking plan class 'BRP6-cuda32-OSX'
2016-01-19 06:31:06.1037 [PID=12895] [version] parsed project prefs setting 'gpu_util_brp': 0.000000
2016-01-19 06:31:06.1037 [PID=12895] [version] No CUDA devices found
2016-01-19 06:31:06.1037 [PID=12895] [version] Checking plan class 'BRP6-opencl-ati-lion'
2016-01-19 06:31:06.1037 [PID=12895] [version] OS version required max: 140000, supplied: 150200
2016-01-19 06:31:06.1037 [PID=12895] [version] no app version available: APP#29 (einsteinbinary_BRP6) PLATFORM#10 (x86_64-apple-darwin) min_version 0
2016-01-19 06:31:06.1037 [PID=12895] [version] no app version available: APP#29 (einsteinbinary_BRP6) PLATFORM#6 (i686-apple-darwin) min_version 0
2016-01-19 06:31:06.1086 [PID=12895] [mixed] sending locality work second
2016-01-19 06:31:06.1114 [PID=12895] [debug] [HOST#2242651] MSG(high) No work sent
2016-01-19 06:31:06.1114 [PID=12895] [debug] [HOST#2242651] MSG(high) see scheduler log messages on https://einsteinathome.org/host_sched_logs/2242/2242651
2016-01-19 06:31:06.1114 [PID=12895] [debug] [HOST#2242651] MSG(high) Binary Radio Pulsar Search (Arecibo) is not available for your type of computer.
2016-01-19 06:31:06.1115 [PID=12895] Sending reply to [HOST#2242651]: 0 results, delay req 60.00
2016-01-19 06:31:06.1116 [PID=12895] Scheduler ran 0.103 seconds
No new Work units for over a week
)
You're only requesting GPU work and it gets denied because of the following line in the log:
Someone with a Mac might be able to help with this problem, but from the looks of it I'd say a downgrade of the OS might be required, or maybe there is a beta app that's allowed on 15.2?
Holmis: Thanks for this
)
Holmis:
Thanks for this info. I can't figure why I am requesting GPU work only. I usually get all kinds of CPU work.
I'll see if I can figure out how it went to GPU only.
/r
Curtis
Your tasks list only shows
)
Your tasks list only shows CPU tasks (Gamma ray pulsar - FGRP4) so it doesn't look like you were getting GPU tasks previously. This snippet from the log
clearly shows that your BOINC client was not asking for CPU work (0.00 secs) but was asking for GPU work (86400 secs worth). Your version of OS X seems to be preventing you from getting GPU work. It might be worthwhile seeing if changing your EAH project preferences to allow beta test apps might override the OS X version restriction. I vaguely remember seeing something of this nature at one point so it's probably worth a try. I don't really know anything about the OS X apps.
It would appear that you must have changed your preferences to stop requests for CPU work. Check your EAH preferences to see if CPU tasks are allowed and that you have selected the Gamma ray pulsar search #4 science run. If you allow your machine to ask for work, there should be no reason why you wouldn't get some. There's no shortage that I know of.
Have you upgraded the OS in your machine recently? If so, could that be associated with your current failure to get work?
Cheers,
Gary.
Or it might be that the CPU
)
Or it might be that the CPU cache is full of work from another project, clicking around led me to MilkyWay where multiple CPU tasks are listed for the same host.
Good point!! I must
)
Good point!!
I must remember that a low RAC at another project doesn't mean a low resource share or a lack of onboard work for that project.
The aborted Einstein tasks due to deadline issues points to the machine not having a high uptime. I guess the machine was in panic mode trying to get those tasks done so now it's going to favour other projects for a while until the debt is repaid.
Cheers,
Gary.
Gary: Thanks for you help -
)
Gary:
Thanks for you help - much appreciated.
No I haven't updated the major OS X since EL Capitan came out last Fall. It's only been the occasional OS bug /security update since then??
This was working fine then it just stopped accepting jobs. The only thing that I recall happening was that I had some EAH tasks that just would not complete. The countdown clock keep resetting on them so I eventually aborted them. There were 3 or 4 of them as I recall. That happened after I updated to the latest BOINC release. (I had everything stopped when I installed the new BOINC version). After that I could not get tasks again. What worries me a bit is the log entry that says "unreliable" and it points to a reliability rate of .09 > .05 - which I think is the number of task failures. Is it possible that I somehow clipped a task failure threshold. I've been running tasks for years and have completed thousands so far.
Also, on the question of CPU queue being full with Milkyway tasks. I issued a NO NEW TASKS to MilkyWay and Asteroids yesterday and the CPU queue is now down to only 8 tasks running and waiting. No improvement in getting Einstein tasks. EAH just doesn't like me anymore.......
After all the tasks run to completion, I am thinking of reverting back to the previous BOINC version if Einstein still won't give me tasks. At lest as an experiment.
I will also try accepting Beta's and see if that helps.
Thanks for helping. I'd like to get Einstein running again.
/r
Curtis
If you are not getting any
)
If you are not getting any new work for FGRP4 - while it lasts, as it´s at less than 7.5% remaining right now - it usually means one or several of the following:
- Einstein has been set to NNT (no new tasks);
- another project is eating up all available resources, possibly in panic mode.
That´s all assuming that nothing is wrong with the host. Our systems are not comparable, but I´d be surprised with yours having a problem seeing that everything except Einstein at Home runs fine. For what it´s worth, I am crushing almost exclusively FGRP4, no issues at all due to OS X upgrades, latest being this week´s 10.11.3 release (Darwin 15.3 - El Capitan). However, that´s on a quite old iMac coming initially with Leopard, all updates applied since. I don´t have a suitable GPU either, of course.
Allowing beta test apps has
)
Allowing beta test apps has certainly worked to get some GPU tasks. Currently you have 4 and one has been completed and validated.
Your BOINC client is still NOT asking for CPU work. The quickest way to see if it is allowed (by preferences) to ask, is to temporarily suspend ALL CPU tasks for other projects to see if this will force a request for Einstein CPU tasks. IF that works, then you know the reason BOINC is NOT asking for CPU tasks is because it thinks other projects are more deserving. In that case you should resume all suspended tasks and let BOINC get on with its job of managing things according to your resource share settings.
There were 4 that were aborted and I think 3 of them may have been close to or even past the deadline. I wondered if you had just aborted them for that reason. I'm not sure what you mean by "countdown clock keep resetting". Could you explain what was happening? FGRP4 tasks progress in 'steps' rather than continuous increments every second. That behaviour, although quite normal, can make it look like nothing is happening.
I have no experience with the latest BOINC versions. All my hosts run Linux and I'm running 7.2.42 which is still the recommended version for Linux. Perhaps there are changes in how 7.6.22 works out which projects should next fetch work but that wouldn't be any sort of permanent block on Einstein. If you can get CPU tasks by the above method, then just let BOINC do its job without worrying about it.
My guess is that this is of no concern to you. I don't know exactly what it means but my guess it's to do with projects that might be using very high 'reliability' to allow only one task in a quorum - ie. no 2nd task needed to validate the 'correctness' of a result. It's not used for this at Einstein and it wouldn't be the reason for not getting CPU tasks.
If BOINC doesn't want to get Einstein at the moment, I don't think just setting NNT on other projects will work, UNLESS there is nothing else to crunch.
I think you should persevere with the current version for a while yet.
There's actually a more serious problem for you to look into. If you look at the remaining FGRP4 CPU tasks that show up on the website, you can see a huge difference between CPU time and elapsed time for all results. Such large differences are not normal. You have a quad core CPU with HT - 8 virtual cores and BOINC sees the full 8. Is your machine an iMac or a Macbook? Are you running CPU tasks on all 8? Are you monitoring CPU temperatures? Apart from BOINC stuff, are you running other CPU intensive work?
Cheers,
Gary.
It's now up to 8 GPU tasks of
)
It's now up to 8 GPU tasks of which 5 have been returned.
Your RAC will be going through the roof! :-).
Cheers,
Gary.
RE: Thanks for helping. I'd
)
I've just posted a response to a similar problem in this new thread. You should read the opening post and my response and also look at a message on the BOINC boards that I linked to.
I give this advice since your recent successful entry into the Einstein GPU crunching market is likely to further reduce your chances of getting more CPU tasks here. To understand why, you need to follow the link and read it all carefully. It may be much more than you really want to digest :-).
Cheers,
Gary.