Development for LHC@home

Well, one immediate difference, the directory started out at ~1.2 GB in Mint Linux, but 7.4 GB in Windows 7! There's a 5.8 GB vm_cache.vdi in Windows that doesn't exist on Linux:

Hi ivan,

It looks like you have a local problem. That vm_cache.vdi shouldn't be there.
That's a remnant from an old task (you see it's from 19 March). The vdi is now called vm_image.vdi.
If you have a new task, you will see both: vm_image.vdi and that old vm_cache.vdi. You can remove that old file.

ps: after 16½ hours running the slot-dir on my Win-system now has a size of 4891 MB.

Well, one immediate difference, the directory started out at ~1.2 GB in Mint Linux, but 7.4 GB in Windows 7! There's a 5.8 GB vm_cache.vdi in Windows that doesn't exist on Linux:

Hi ivan,

It looks like you have a local problem. That vm_cache.vdi shouldn't be there.
That's a remnant from an old task (you see it's from 19 March). The vdi is now called vm_image.vdi.
If you have a new task, you will see both: vm_image.vdi and that old vm_cache.vdi. You can remove that old file.

ps: after 16½ hours running the slot-dir on my Win-system now has a size of 4891 MB.

Yes, I've just noticed that this morning -- it was still in the directory.
So, panic over I think, but it's still something to keep an eye on.
I'll clean up and start another task when I get into work.

---

All 4 CMS tasks (2 Win 2 Linux) still OK here.

Longest running Win task @ 5 1/4 hrs,
Project dir.......1.4 GB
Slot dir..........5.0 GB of which snapshot dir is 487MB.

All 4 CMS tasks (2 Win 2 Linux) still OK here.

Longest running Win task @ 5 1/4 hrs,
Project dir.......1.4 GB
Slot dir..........5.0 GB of which snapshot dir is 487MB.

Thanks, I'll have the data to make some plots tomorrow^Wlater today (24-hour experiments take some time to complete...), but it's obvious now that my initial problem was that large file somehow left behind from an aborted task -- and subsequent tasks landing in that same slots directory! However, the increase in disk size is something we'll have to watch carefully as we go into beta and thence into production.

---

Ivan,

We've been concentrating on Win, but the increase in
disk use also occurs on Linux.

The longest running Linux task here is:-

Running time....8hrs
Project dir.....1.4GB
Slot dir........6.3GB of which snapshot dir is 404MB.

I'll set NNW as requested.

John.

This
task failed on restart after the host had been shut down. This
occasionally happens on other projects. Currently downloading new
version 26158.
What VBox version must I use for this to work? Currently still on 4.3.12
which failed the new BOINC task on host shutdown (from GUI via VNC)

We've been concentrating on Win, but the increase in
disk use also occurs on Linux.

I think that's because the VM accumulates the results of each "job" it runs. At the moment I'm not completely clear on when results are returned to the Data Bridge, and if they're deleted from the VM when/if they are.
Ultimately the problem I had was an image file left undeleted from a previous (probably aborted) run.

---

What VBox version must I use for this to work? Currently still on 4.3.12
which failed the new BOINC task on host shutdown (from GUI via VNC)

As far as I can tell, we can all now run 4.3.26 with the latest VBoxWrapper -- please let us know if that's not the case!

---

...
However, the increase in disk size is something we'll have to watch carefully as we go into beta and thence into production.

My current task (running with a vboxwrapper not making snapshots) has used 6061 MB in the slot directory after almost 16 hours running.

What VBox version must I use for this to work? Currently still on 4.3.12
which failed the new BOINC task on host shutdown (from GUI via VNC)

As far as I can tell, we can all now run 4.3.26 with the latest VBoxWrapper -- please let us know if that's not the case!

OK, I've changed all hosts running CMS to VBox 4.3.26-98988. Still finishing off a couple of 26158 tasks - running OK on Win and Linux.

.... However, the increase in disk size is something we'll have to watch carefully as we go into beta and thence into production.

I also had a vm_image.vdi as a leftover in a slot directory from a VM-task, producing errors for another project:

SRBase 31 Mar 19:18:29 Aborting task R79_43-45k_wu_331790_0: exceeded disk limit: 5120.13MB > 953.67MB

Probably caused by a task cancelled by the project: http://boincai05.cern.ch/CMS-dev/result.php?resultid=35155