723386 - reserved slaves feature is broken

Status: RESOLVED FIXED

(Release Engineering :: Release Automation: Other, defect, P4)

Description

[Rail Aliiev [:rail]]

I set it to 10 this morning and after 10-12 hours still have pending jobs :/

Need some love. :)

Comment 1

[Rail Aliiev [:rail]]

Haven't touched this for a long time, back to the pool

Comment 2

[Rail Aliiev [:rail]]

Tested in staging. The master doesn't start _new_ builds. Probably I was hit by long running builds...

Comment 3

[Chris AtLee [:catlee]]

pretty sure this is still happening

bm08 has reserved slaves set to 6, and has 5 win32 ix slaves attached. 4/5 are running non-release builds at the moment even though release jobs are pending.

Comment 4

[Chris AtLee [:catlee]]

specifically, the following builders are running:

mozilla-inbound_win32_spidermonkey-warnaserr
WINNT 5.2 mozilla-inbound leak test build
mozilla-inbound_win32-debug_spidermonkey-nomethodjit 
mozilla-inbound_win32-debug_spidermonkey-warnaserrdebug

we're currently pending on the repack jobs

Comment 5

[Chris AtLee [:catlee]]

I believe the problem is that since we've removed the windows VMs, all windows machines are now considered 'slow'.

The reserved slaves file only reserves 'fast' slaves.

Comment 6

[Chris AtLee [:catlee]]

Attachment: fixes for slave reservations

this removes the distinction between reserved fast and reserved slow slaves, which we never used anyway.

now reservations will reserve fast slaves first, and then reserve as many slow slaves as required to fulfill the reservation. If there are no fast slaves, then only slow slaves will be reserved.

NB that for windows and macosx64 *all* slaves are considered slow here.

Comment 7

[Chris AtLee [:catlee]]

deployed as part of this morning's reconfig

Comment 8

[[e-mail] (:bhearsum)]

Bulk move of bugs to Release Automation component.

Comment 9

[Chris AtLee [:catlee]]

Ben reports this is working for the win32 repacks for the latest beta.