Skip to content

Crash rollout engines from the soak harness instead of the controller - #2120

Open
fzyzcjy wants to merge 1 commit into
tom/refactor-miles/op13-127from
tom/refactor-miles/op13-112
Open

Crash rollout engines from the soak harness instead of the controller#2120
fzyzcjy wants to merge 1 commit into
tom/refactor-miles/op13-127from
tom/refactor-miles/op13-112

Conversation

@fzyzcjy

@fzyzcjy fzyzcjy commented Aug 3, 2026

Copy link
Copy Markdown
Collaborator

Part of #1837

@gemini-code-assist

Copy link
Copy Markdown
Contributor

Caution

The consumer version of Gemini Code Assist on GitHub has been sunset. All code review activity has officially ceased.

@fzyzcjy
fzyzcjy force-pushed the tom/refactor-miles/op13-112 branch 4 times, most recently from f14c938 to 7666c0c Compare August 4, 2026 02:36
@fzyzcjy
fzyzcjy force-pushed the tom/refactor-miles/op13-127 branch from 658b39c to 5ba57d0 Compare August 4, 2026 02:36
@fzyzcjy
fzyzcjy force-pushed the tom/refactor-miles/op13-112 branch from 7666c0c to f7bfdf4 Compare August 4, 2026 06:53
@fzyzcjy
fzyzcjy force-pushed the tom/refactor-miles/op13-127 branch from 5ba57d0 to ce2795f Compare August 4, 2026 06:53
@fzyzcjy
fzyzcjy force-pushed the tom/refactor-miles/op13-112 branch 2 times, most recently from a88efd2 to a97d34d Compare August 4, 2026 07:31
@fzyzcjy
fzyzcjy force-pushed the tom/refactor-miles/op13-127 branch from ce2795f to dbfdd32 Compare August 4, 2026 07:58
@fzyzcjy
fzyzcjy force-pushed the tom/refactor-miles/op13-112 branch from a97d34d to 9123e88 Compare August 6, 2026 03:46
@fzyzcjy
fzyzcjy force-pushed the tom/refactor-miles/op13-127 branch from dbfdd32 to 0521fa7 Compare August 6, 2026 03:46
@fzyzcjy
fzyzcjy force-pushed the tom/refactor-miles/op13-112 branch from 9123e88 to 6a7a947 Compare August 6, 2026 07:04
@fzyzcjy
fzyzcjy force-pushed the tom/refactor-miles/op13-127 branch from 0521fa7 to 2a12380 Compare August 6, 2026 07:04
@fzyzcjy
fzyzcjy force-pushed the tom/refactor-miles/op13-112 branch from 6a7a947 to 308c241 Compare August 9, 2026 23:31
@fzyzcjy
fzyzcjy force-pushed the tom/refactor-miles/op13-127 branch 2 times, most recently from c062466 to cf6b367 Compare August 10, 2026 00:44
@fzyzcjy
fzyzcjy force-pushed the tom/refactor-miles/op13-112 branch 2 times, most recently from 68cd73d to 318bd70 Compare August 10, 2026 06:00
@fzyzcjy
fzyzcjy force-pushed the tom/refactor-miles/op13-127 branch from cf6b367 to dcffe00 Compare August 10, 2026 06:00
@fzyzcjy
fzyzcjy force-pushed the tom/refactor-miles/op13-127 branch from dcffe00 to a7917f2 Compare August 10, 2026 16:06
@fzyzcjy
fzyzcjy force-pushed the tom/refactor-miles/op13-112 branch 2 times, most recently from 01505ee to ad75a46 Compare August 10, 2026 16:33
@fzyzcjy
fzyzcjy force-pushed the tom/refactor-miles/op13-127 branch from 0ac3425 to fdbd7af Compare August 12, 2026 02:48
@fzyzcjy
fzyzcjy force-pushed the tom/refactor-miles/op13-112 branch 2 times, most recently from e645ff3 to cac8e31 Compare August 12, 2026 09:32
@fzyzcjy
fzyzcjy force-pushed the tom/refactor-miles/op13-127 branch from 4dc05ec to c120733 Compare August 12, 2026 14:31
@fzyzcjy
fzyzcjy force-pushed the tom/refactor-miles/op13-112 branch from cac8e31 to 318a3b5 Compare August 12, 2026 14:31
The controller injected CI faults into itself, which needed it to reach into its
own engines and to know it was under test. Faults are now injected the way
trainer ones already are, over the api server, so the controller keeps no test
hook at all.

The colocated mode this adds is the most direct evidence that colocate x ft
works: an engine is crashed mid training, relaunched gated, and only comes back
to Running at the next weight update window.
@fzyzcjy
fzyzcjy force-pushed the tom/refactor-miles/op13-112 branch from 318a3b5 to a113d13 Compare August 12, 2026 14:36
@fzyzcjy
fzyzcjy force-pushed the tom/refactor-miles/op13-127 branch from c120733 to 5620b7f Compare August 12, 2026 14:36
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant