Closed Bug 524820 Opened 10 years ago Closed 10 years ago

purge_builds.py needs to know about scratchbox

Categories

(Release Engineering :: General, defect, P2)

defect

Tracking

(Not tracked)

RESOLVED FIXED

People

(Reporter: aki, Assigned: coop)

References

Details

Attachments

(2 files, 1 obsolete file)

Either we go with the bind mount fix which should be transparent, or purge_builds.py needs to know how to account for and clean up scratchbox.
Blocks: 522719
For reference, when bug 519684 and bug 524010 land there will be left overs on scratchbox since their build directory names are going to change.

I will manually remove things with a lot of care.
Taking since I am afraid that the build dirs renaming for mobile means more code checked out and we might run out of space on linux slaves.
Assignee: nobody → armenzg
Status: NEW → ASSIGNED
Blocks: 519684
John Ford is going to be taken this from me so I can dedicate myself more on other bugs.

John feel free to use my WIP or take your own approach.

Thanks tons!
Assignee: armenzg → jford
Unassigning; My bad, John understood look at the patches not actually taking the bug and he has his hands full with buildduty tasks.
Let's see who can take on this.
Assignee: jford → nobody
This has lingered long enough. I'll try to put it out of its misery tomorrow.
Assignee: nobody → ccooper
OS: Linux → All
Priority: -- → P2
Hardware: x86 → All
Attachment #410034 - Flags: review?(aki) → review+
We only really need to add this dir on linux, but because of the inheritance patterns of our factories, not all of them have something as easy as self.platform to reference to see if we should be adding it or not. Easier to just add it across the board and have the path not be found (non-fatally I might add, it just cycles to the next base_dir) on mac/win32.

Tested successfully in staging with the purge_builds.py patch this afternoon.
Attachment #410047 - Flags: review?(aki)
Comment on attachment 410047 [details] [diff] [review]
Add scratchbox build dir to list of dirs to purge

Right, it does do an os.path.exists on each base_dir so that shouldn't be an issue.

I am a bit concerned with whether this will cause confusion ("why are we trying to purge scratchbox on a windows box?"), but since disk space issues are killing us right now, the tradeoff works.
Attachment #410047 - Flags: review?(aki) → review+
(In reply to comment #9)
> I am a bit concerned with whether this will cause confusion ("why are we trying
> to purge scratchbox on a windows box?"), but since disk space issues are
> killing us right now, the tradeoff works.

I'll add an explanatory comment to factory.py when I land.
Comment on attachment 410034 [details] [diff] [review]
Allow purge_builds.py to accept multiple base_dirs for purging

http://hg.mozilla.org/build/tools/rev/82e039d366f6
Attachment #410034 - Flags: checked-in+
Comment on attachment 410047 [details] [diff] [review]
Add scratchbox build dir to list of dirs to purge

http://hg.mozilla.org/build/buildbotcustom/rev/a75e62a5e844
Attachment #410047 - Flags: checked-in+
Reconfig-ed both production-masters with this change tonight.
Status: ASSIGNED → RESOLVED
Closed: 10 years ago
Resolution: --- → FIXED
Do we also need to clean out /scratchbox/users/cltbld/home/cltbld/l10n?
(In reply to comment #14)
> Do we also need to clean out /scratchbox/users/cltbld/home/cltbld/l10n?
>
Everything is going to fall under build/ when bug 524010 lands.
(In reply to comment #14)
> Do we also need to clean out /scratchbox/users/cltbld/home/cltbld/l10n?

Armen had indicated to me that he was going to do that by hand when bug 524010 lands. csshX is his friend.
Product: mozilla.org → Release Engineering
You need to log in before you can comment on or make changes to this bug.