793170 - mountain-lion talos systems using the wrong basedir

Status: RESOLVED INCOMPLETE

(Infrastructure & Operations Graveyard :: CIDuty, task)

Description

[Dustin J. Mitchell [:dustin] (he/him)]

/tools/buildbot-0.8.4-pre-moz2/bin/python2.7 /tools/buildbot/bin/twistd --no_save --logfile /builds/slave/talos-slave/twistd.log --python /builds/slave/talos-slave/buildbot.tac

That should be /builds/slave/buildbot.tac like other slaves.  I'm not sure how/where that went wrong, but we should get it fixed for consistency with other slaves.

Comment 1

[Kim Moir [:kmoir] ET]

I've fixed this in  slavealloc
update slaves set basedir="/builds/slave/" where name like "%mtnlion%";

Perhaps I should wait until they have all rebooted and then delete the old buildbot.tac via cssh

Comment 2

[Dustin J. Mitchell [:dustin] (he/him)]

A puppet patch could take care of that more easily?

Comment 3

[Kim Moir [:kmoir] ET]

I reverted this change because I was talking to dustin in irc realized this would impact the Apache document root etc.

and he mentioned that for puppet again he was hoping
dustin	I had hoped to change ~cltbld/talos-slave to /build/slave
dustin	but it got changed to /build/slave/talos-slave instead
dustin	(because if we can change it to /builds/slave, then at some point there's no need for a conditional anymore)

Comment 4

[Kim Moir [:kmoir] ET]

Attachment: patch to delete old buildbot.tac

Comment 5

[Dustin J. Mitchell [:dustin] (he/him)]

So, here's a proposed way to get this fixed:

In puppet,
 1. remove /builds/slave/talos-slave and make it a symlink to /builds/slave
 2. change apache docroot to omit "/talos-slave"
Step 1 will probably need an exec {} - IIRC file {} won't replace a dir with a symlink.

Then, once that's active on the masters, make the change in comment 1 in slavealloc.

After a few weeks, we can remove step 1.

Thoughts?

Comment 6

[Dustin J. Mitchell [:dustin] (he/him)]

The new linux testers have this incorrect basedir, too (bug 861200).

Comment 8

[Justin Wood (:Callek)]

I think we're done here. Kim?

Comment 9

[Dustin J. Mitchell [:dustin] (he/him)]

Not at all - nothing's been changed.

Comment 10

[Kim Moir [:kmoir] ET]

No, I didn't implement this change.

Comment 11

[Justin Wood (:Callek)]

last functional change on this bug was September, I am tossing it back to the buildduty pool. It might better belong in General Automation or Platform Support however.

Comment 12

[Dustin J. Mitchell [:dustin] (he/him)]

This was more of a "let's make this right before we replicate this mistake everywhere" thing.  We've now replicated the mistake everywhere.  I don't think it's worth fixing at this point.