637973 - Re-add mv-moz2-linux-ix-slave22 to production_config.py and get it back in service

Status: RESOLVED FIXED

(Release Engineering :: General, defect, P3)

Description

[John O'Duinn [:joduinn] (please use "needinfo?" flag)]

We scavenged a few unused win64 ix machines for use as buildbot masters in SCL1. This means we currently have no w64-ix-slave{01...06}. 

Now that we're getting closer to the win64 ref image being ready to roll out, we should fill out the win64 rack. Zandr and I talked, and propose to:
* take 6 of the least used IX machines in 650castro that are due to be moved soon anyway (exact machines TBD with RelEng buildduty).
* re-image & rename them to w64-ix-slave{01...06}
* move these into the empty slots in rack in SCL1, beside w64-ix-slave07.b.m.o

Comment 1

[Armen [:armenzg]]

These machines are not connected to production systems so feel free to tackle them. There is no way it could affect any systems. Sorry if it wasn't explicit.

RC1 might be due tomorrow so another embargo could happen tomorrow.

Comment 2

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

these are from the set of currently-disabled slaves in mtv:

mv-moz2-linux-ix-slave20
mv-moz2-linux-ix-slave21
mv-moz2-linux-ix-slave22
mw32-ix-slave19
mw32-ix-slave20
mw32-ix-slave21

Comment 3

[John O'Duinn [:joduinn] (please use "needinfo?" flag)]

zandr thinks he'll be in scl later today, in which case he will move these as they are already down.

Comment 4

[Zandr Milewski [:zandr]]

Do I need to do anything special when I turn these back on in scl1, or are they down in such a way that they're "safe"?

Comment 5

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

they are down in such a way that they're "safe"

Comment 6

[Zandr Milewski [:zandr]]

The move isn't going to happen today, unscrewing the AFK ratsnest took a long time.

Comment 7

[Zandr Milewski [:zandr]]

Can't actually do this until bug 645024 is resolved. Reversing the dependency.

Comment 8

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

It looks like all six of these have been pressed back into their original purposes - the three w32 boxes are running in try, and the linux boxes are doing linux boxy things.

So, I'm officially aborting this operation.  If we come to need more w64 builders later, we can open up a new bug.

However, somehow of the six hosts, only mv-moz2-linux-ix-slave22 is not in production_config.py.  So back into the releng pool to get that re-added and the reconfig done.

Comment 9

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

production_config.py changes aren't a slaveduty thing

Comment 10

[Nick Thomas [:nthomas] (UTC+12)]

mv-moz2-linux-ix-slave22 is actually in the list for try ix's:
 TRY_LINUX_IXS  = ['mv-moz2-linux-ix-slave%02i' % x for x in range(22,24)] + \
but in slavealloc was set to Trust: core, Pool: build-sjc1. Given the existing config, history of try jobs, and try ssh keys, I've corrected those fields (try, try-mtv1) and rebooted it. Puppet went fine.