1886329 - test out smaller worker types for android emulator tests

Status: RESOLVED FIXED

(Testing :: General, task)

Description

[Joel Maher ( :jmaher ) (UTC -8)]

we run on n2-standard-16, but we only run a single instance on that; possibly we can run on n2-standard-8 or n2-standard-4; this would save cost.

Comment 1

[Joel Maher ( :jmaher ) (UTC -8)]

Attachment: Bug 1886329 - add android emulator t-linux-kvm-gcp staging pools for worker exploration. r=aerickson

Comment 2

[Joel Maher ( :jmaher ) (UTC -8)]

leave-open as we might adjust a bit more and then we need to cleanup

Comment 3

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/ci/ci-configuration/rev/9c7504501ba2
add android emulator t-linux-kvm-gcp staging pools for worker exploration. r=aerickson

Comment 4

[Joel Maher ( :jmaher ) (UTC -8)]

Attachment: Bug 1886329 - fix unittest to temporarily allow min_disks for staging pool. r=bhearsum

Comment 5

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/ci/ci-configuration/rev/bb543bb2bed3
fix unittest to temporarily allow min_disks for staging pool. r=bhearsum

Comment 6

[Joel Maher ( :jmaher ) (UTC -8)]

Attachment: Bug 1886329 - add more kvm worker pools to test android-em. r=aerickson

Comment 7

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/ci/ci-configuration/rev/fe01862c2cac
add more kvm worker pools to test android-em. r=aerickson

Comment 8

[Joel Maher ( :jmaher ) (UTC -8)]

Attachment: Bug 1886329 - remove pool w/reference to persistent-disk-60 for now. r=aerickson

Comment 9

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/ci/ci-configuration/rev/a2f9d586bb71
remove pool w/reference to persistent-disk-60 for now. r=aerickson

Comment 10

[Joel Maher ( :jmaher ) (UTC -8)]

Attachment: Bug 1886329 - remove min_disk hack as testing with fewer disks. r=aerickson

Comment 11

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/ci/ci-configuration/rev/efa16879012f
remove min_disk hack as testing with fewer disks. r=aerickson

Comment 12

[Joel Maher ( :jmaher ) (UTC -8)]

Attachment: Bug 1886329 - remove invalid '.' in worker pool id. r=aerickson

Comment 13

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/ci/ci-configuration/rev/6ba12636892b
remove invalid '.' in worker pool id. r=aerickson

Comment 14

[Joel Maher ( :jmaher ) (UTC -8)]

Attachment: Bug 1886329 - test out no disk for linux and android, also different builder sizes for ac-android. r=aerickson

Comment 15

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/ci/ci-configuration/rev/a159a66ca71d
test out no disk for linux and android, also different builder sizes for ac-android. r=aerickson

Comment 16

[Joel Maher ( :jmaher ) (UTC -8)]

Attachment: Bug 1886329 - remove 40GB disks and keep 75GB disks for now. r=aerickson

Comment 17

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/ci/ci-configuration/rev/9e573819accd
remove 40GB disks and keep 75GB disks for now. r=aerickson

Comment 18

[Joel Maher ( :jmaher ) (UTC -8)]

Attachment: Bug 1886329 - increase the pool size of nodisk workers to gather metrics. r=aerickson

Comment 19

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/ci/ci-configuration/rev/b2bb07b9d866
increase the pool size of nodisk workers to gather metrics. r=aerickson

Comment 20

[Joel Maher ( :jmaher ) (UTC -8)]

many experiments here, so far the outcomes are:

deployed:

• reduced t-linux-kvm-gcp from n2-standard-16 -> n2-standard-8 (half price) and from 2 scratch-disk -> 1

need to write patch:

• create builder for ac-android-all which does n2-standard-16 -> n2-standard-8 (half price)
• remove scratch-disk from t-linux-kvm-gcp
• remove scratch-disk from t-linux-large-gcp

need to finish experiment:

• remove scratch-disk from t-linux-xlarge-gcp (waiting on ccov)

further experiments will be:

• reducing disk size from 75GB -> 40GB for t-linux/kvm
• reducing tasks that use t-linux-xlarge-gcp -> t-linux-large-gcp

Comment 21

[Joel Maher ( :jmaher ) (UTC -8)]

I have done all this work and cleaned up what is needed in worker-pool.yml; There are more experiments to do, but for now I would like to mark this specific bug done.

what remains from earlier list is:

• reducing disk size from 75GB -> 40GB for t-linux/kvm
• reducing tasks that use t-linux-xlarge-gcp -> t-linux-large-gcp

possibly test out c2-* workers; maybe adjust ram;