Labour intensive doesn't begin to describe this.... and my hub died last night ( dead as in no wifi, no LAN, no resurrection from reboots, resets nada. All lights on but no one home. It would appear to be a dead parrot ) more wasted hours
The best form of help from above is a sniper on the rooftop....
I keep suffering from units sitting idle with 'Waiting to run (Scheduler wait: VM job unmanageable, restarting later)'
Shutting down the client and restarting seems to start the jobs off and running for a while.
Is anyone else seeing this behavior ? Have found a few references to it with a web search but no explanation of what it is or why it is being caused.
The best form of help from above is a sniper on the rooftop....
Yup I've had a few such messages. Restarting the client seems to be the only hope although it sometimes kills the WU. On the other hand just looking at the VM thingy invariably kills the WU.
Despite that & your hardware issues (I see you are moving again - well done) we are still 6K ahead of our Italian rivals which is a useful but not yet conclusive margin.
What version BOINC and virtualbox you running? I have slightly different versions on a few, but the ones running BOINC 7.2.42 and virtualbox 4.3.12 seem to be most stable on Win7 64 and Win8 64 systems. I have two 3 core AMD systems that won't run the 2nd instance correctly. It launches the task, but the % complete never updates and never finishes correctly. I had an issue like yours on that system recently while trying to run the 2nd task. Had to suspend T4T on both clients, shut down both clients, make sure vbox didn't have anything running (may have had to power off, remove instance with the files), then restart the 2nd client and resume the task. Not sure if started from scratch because the % complete won't update. I'm just hoping it does. I've decided that once a system has the number of clients running, leave it alone except to check on and update the project if the task completed and needs to be uploaded (else it's sitting there doing nothing with it's single completed task).
Other than that, make sure the client computing preferences are set to use while computer is in use, while processor usage is less than 0 (0 means ingore) and on multiprocessor systems, use at most 100%.
On one system, I had to comlpetely uninstall the older version of vbox, including folder /users/<user name>/.virtualbox, reboot, then install 4.3.12.
Must have had a premonition of impending doom as I bought a new homehub 4 a couple of weeks ago with the intention of replacing the now dead hub and keeping it as a backup. Guess I will have to get another backup now
All my hosts are on Boinc 7.3.15 running vbox 4.3.12.
Seems to be a case of once a unit has a problem it just keeps on giving. This time I tried shutting down all the hosts on the machine and boinc then firing them all back up again. See if this makes any difference.
I don't use the vm manager to check the units but connect to their respective boinc manager and check them that way.
The best form of help from above is a sniper on the rooftop....
Well, we managed to keep our lead today. I hope we have some WUs coming in soon. Those pesky Italians are right on our tail.
I think there might be a strategy for managing clients for their finals WU's. The WUs take around take 24 hours to complete and though I've not really studied this, I think the more CPU time they get, the more credit you may get. Based on this assumption and the fact I'm wore out running this project and don't want any clients downloading any WU that I'm sure won't finish before the project ends, it's probably worth a few more credits to stage when to stop allowing new tasks. I think it's probably common sense to do this for you veterans, but this is an odd project that takes 24 hours to complete a WU. Just pointing it out.
With just over 24hr to go we are 10K ahead which should probably be enough. :)
But it will be close.
There doesn't seem to be a 24hr limit on WUs (I've got one which has been running for 49 hrs) but nothing returns in less than 24 hrs. I agree that there is no point in forcing new clients and we should concentrate on keeping the running ones alive.
There seems to be a pattern that TSBT do well over-night but BOINC.Italy pick up credits during the day with us gaining ground later in the evening. So it will be an exciting watch but I'm quietly confident that we will pull this off! :)
Unfortunately due to my hubs suicide I will have units that will upload 1 - 2 hours after the comp finishes. Have come to the conclusion that once a unit starts erroring it will keep doing so and it seems best to just cut your losses and ditch it.
I haven't had a single unit run longer than 24hrs , are you sure the 49 hr unit is running ? I have found that the boinc client is best to check as the vm manager reports units running that are actually dead when you check them in boinc.
This could still be a close run thing but hopefully we can pull this off
The best form of help from above is a sniper on the rooftop....
5 hours 21 minutes to go. They're still 12,000 behind, which is probably far enough, but I hope we have few more WU that will finish. I have 3,000-4,000 worth that I'm pretty sure will finish before then. Anyone else have any that will finish in time?
Well done everyone, some impressive numbers being generated by you boys up at the top
Hopefully I'll have time to try multiple clients on the next comp :)