Maximum disk usage exceeded

Ray Murray
Ray Murray
Joined: 22 Jan 05
Posts: 3
Credit: 119655
RAC: 0

Not been here for quite a

Not been here for quite a while since concentrating on the various Cern projects from where I have come in reply to Ritterm's post. With there being a few threads on this subject across various projects, I don't want to clutter the boards with multiple copied posts but I hope these observations can be passed to those that need them.

The failure to delete the CMS-dev .vdi is repeatable every time on my 2 hosts. I've not had many CMS failures but even successful tasks leave the image in the slot.
Everything else in the slot gets deleted and the VM is removed from VBox but the .vdi remains in the slot. If CMS-dev is next to use that slot, the debris image is replaced by the new image. If another project eg. LHC(Sixtrack) tries to use that slot, the 4-5GB image left there is usually enough to send it over limit resulting in error.

A restart of Boinc doesn't clear it. I've not tried a reboot but don't think it would help.

The offending image can be manually deleted straight away, freeing up the slot for immediate use by ANY project.

If Boinc is thinking the image is still in use, would a short delay help. The VM is removed from VBox so perhaps a delay there would allow VBox to tidy up THEN let Boinc tidy up.
I've not caught CMS "in the act" of finishing but I seem to recall VLHC/T4T stops the VM and clears everything out a short while before Boinc reports the task as finished. I'll see if I can catch a CMS one later tonight or suspend one so I can watch the finish tomorrow.

[Edit]
Here's the slot_debug messages for the reuse by CMS of the slot with the previous image left in it:
04/05/2015 20:25:23 | | [slot] cleaning out slots/2: get_free_slot()
04/05/2015 20:25:23 | CMS-dev | [slot] assigning slot 2 to CMS_30670_1427806616.781596_0
04/05/2015 20:25:23 | CMS-dev | [slot] linked ../../projects/boincai05.cern.ch_CMS-dev/vboxwrapper_26165_windows_x86_64.exe to slots/2/vboxwrapper_26165_windows_x86_64.exe
04/05/2015 20:25:23 | CMS-dev | [slot] linked ../../projects/boincai05.cern.ch_CMS-dev/CMS_26_03_2015a.xml to slots/2/vbox_job.xml
04/05/2015 20:26:07 | | [slot] removed file slots/2/init_data.xml
04/05/2015 20:26:07 | CMS-dev | [slot] linked ../../projects/boincai05.cern.ch_CMS-dev/vboxwrapper_26165_windows_x86_64.pdb to slots/2/vboxwrapper_26165_windows_x86_64.pdb

Hoping that is useful in the search for a solution.

Windows 7
Boinc 7.4.42
VBox 4.3.26

I've subscribed to this thread so as to get any further comments or requests for further observations.

Richard Haselgrove
Richard Haselgrove
Joined: 10 Dec 05
Posts: 2139
Credit: 2752648717
RAC: 1486471

Thanks - it'll be useful to

Thanks - it'll be useful to have some experienced watchers to help us get to the bottom of this. I've managed to attach a couple of hosts to CMS-dev, so now I know a little bit more about that to watch for - but I won't get more than a couple of completions per day, so it could be a slow process.

MAGIC Quantum Mechanic
MAGIC Quantum M...
Joined: 18 Jan 05
Posts: 1695
Credit: 1042157711
RAC: 1402731

The vLHC (and CMS-dev) no

The vLHC (and CMS-dev) no longer have the *snapshot for VM* but Atlas is still using the snapshots.

As far as the vdi cache I always check mine to see if they need to have any removed but I tend to have more of those on the one with Win8.1 but I still get some on Win7 but not as many but I use different versions of VB and Boinc so that doesn't seem to be the difference.

But I run all 4 Cern versions and Einstein GPU's at the same time.

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.