Closed Bug 983085 Opened 12 years ago Closed 12 years ago

Deleting user repositories no longer possible

Categories

(Developer Services :: General, task)

x86
All
task
Not set
normal

Tracking

(Not tracked)

RESOLVED FIXED

People

(Reporter: nthomas, Unassigned)

References

Details

Attachments

(1 file)

$ ssh hg.mozilla.org edit tools delete YES $ http://hg.mozilla.org/users/nthomas_mozilla.com/tools is still present. Fallout from local disk hg ?
Does anyone know why this doesn't work? Or have a workaround? I like to refresh my user repos after they get cluttered with test commits, but I can no longer delete/re-clone.
The same here: # fist attempt ssh hg.mozilla.org edit tools delete YES # second attempt ssh hg.mozilla.org edit tools delete YES Could not find the repository at /users/raliiev_mozilla.com/tools. Please check the list at https://hg.mozilla.org/users/raliiev_mozilla.com https://hg.mozilla.org/users/raliiev_mozilla.com is still available and clonable.
(In reply to Rail Aliiev [:rail] from comment #2) > https://hg.mozilla.org/users/raliiev_mozilla.com is still available and > clonable. https://hg.mozilla.org/users/raliiev_mozilla.com/tools I mean
This was related to the migration to having hg served from local disk. The user-repo scripts were never updated to cause a deletion on the webheads. I'm working on fixing this.
As a quick-fix I've deleted the 'tools' repo from both your accounts.
For what it's worth, I was able to delete and reclone my buildbot-configs, buildbotcustom, and tools repositories today without issue.
(In reply to Ben Hearsum [:bhearsum] from comment #6) > For what it's worth, I was able to delete and reclone my buildbot-configs, > buildbotcustom, and tools repositories today without issue. Nevermind, just hit a new symptom. Now I get a different repo depending on where I clone over http or ssh...
Yeah, that's a fun consequence of this. For context, this impacts our ability to test the release automation, where we need to have copies of production repositories for code changes, tagging etc. Lack of deletion makes that more complicated, you either have complicated history in your repo or multiple copies.
As Nick points out in comment 8, this really gets in the way of the releng workflow and slows us down. Based on comment 4 -- can we get an ET on when the script will be fixed, please?
Flags: needinfo?(bkero)
I'm being pulled in at least three directions at the moment, all of them with urgency. I'm basically out the rest of the week, so if I am the only engineering resource dedicated to fixing this then I would say the ET of getting a change in to fix this script would be next week. I can try to put in some work for this script to be fixed on Thursday. This will require changes to hgssh's pash.py script, which has historically taken CAB scheduling and approval, as well as changes to the existing hgweb mirroring script on each hgweb box or the creation of a new script to handle this. Going from the metric of pash.py needing CAB scheduling and approval, so will these changes. Historically they'll also need to be validated by someone in release engineering, which would take more time than the Monday I would have to write them up and hand them off, and since CAB only meets on Tuesday we're looking at around the end of the month (2014-04-29 CAB meeting for beginning-of-May deployment). That said, if you're willing to provide additional engineering resources to make this happen faster, the environment is replicatable using the puppet module available at https://github.com/bkero/puppet-module-hg
Flags: needinfo?(bkero)
Thanks for the quick response -- that gives us the context we need!
I'm attaching he initial version of the changes I've made to the mirror-pull script to allow deletion of repositories (but only user repos)
Looking for a status check here -- who is doing the code review? If these changes are deployed, I believe we're still having issues doing deletions.
Flags: needinfo?(bkero)
Nothing has changed, so it stands to reason the issues doing deletions will be the same. For lack of any code review (did do correctness testing on it) this portion has been deployed. The remaining part is the pash.py script to run the command to issue the deletion across the mirror network, which I hope to finish today.
Flags: needinfo?(bkero)
Blocks: 1016778
ah shucks. I deleted my m-c repo similar to rail in comment 2 http://hg.mozilla.org/users/jlund_mozilla.com/mozilla-central still exists and can be cloned from but I am no longer able to push to it via ssh (the only accepted port to push). I am hoping to either: 1) get a clean copy of m-c in my user repo that works or 2) be able to push to the still visible dirty copy at http://hg.mozilla.org/users/jlund_mozilla.com/mozilla-central bkero, shall I wait for your final fix you mentioned in comment 15 or is there a way I can achieve (1) or (2) myself? I imagine your hands are full and I understand if you're too busy :)
Blocks: 1009807
Flags: needinfo?(bkero)
bkero removed mozilla-central from the mirrors so now it is really gone. I will now create a new clean mozilla-central.
Flags: needinfo?(bkero)
No longer blocks: 1009807
I've pushed the patches out to test this, and can verify that repositories are being deleted correctly. If you're still having an issue with inconsistency please update the bug.
Status: NEW → RESOLVED
Closed: 12 years ago
Resolution: --- → FIXED
Component: Server Operations: Developer Services → General
Product: mozilla.org → Developer Services
You need to log in before you can comment on or make changes to this bug.

Attachment

General

Created:
Updated:
Size: