x/build: OOM on linux-ppc64le-power10osu builder

[gopherbot]

#!watchflakes
post <- builder == "linux-ppc64le-power10osu" && log ~ `signal: killed`

Issue created automatically to collect these failures.

Example (log):

# go run run.go -- rotate1.go
exit status 1
command-line-arguments: /workdir/go/pkg/tool/linux_ppc64le/compile: signal: killed

— watchflakes

[gopherbot]

Found new dashboard test flakes for:

#!watchflakes
post <- pkg == "rotate1.go" && test == ""

2023-01-31 16:54 linux-ppc64le-power10osu go@55a33d88 rotate1.go (log)

# go run run.go -- rotate1.go
exit status 1
command-line-arguments: /workdir/go/pkg/tool/linux_ppc64le/compile: signal: killed

— watchflakes

[gopherbot]

Found new dashboard test flakes for:

#!watchflakes
post <- builder == "linux-ppc64le-power10osu" && log ~ `signal: killed`

2023-01-31 16:53 linux-ppc64le-power10osu go@47e205c3 (log)

FAIL
2023/01/31 17:17:29 Failed: exit status 1
go tool dist: FAILED

2023-01-31 16:53 linux-ppc64le-power10osu go@43115ff0 (log)

FAIL
2023/01/31 17:17:50 Failed: exit status 1
go tool dist: FAILED

2023-02-01 21:30 linux-ppc64le-power10osu go@bd749504 chanlinear.go (log)

# go run run.go -- chanlinear.go
signal: killed

2023-02-01 21:30 linux-ppc64le-power10osu go@cda461bb index0.go (log)

# go run run.go -- index0.go
exit status 1
command-line-arguments: /workdir/go/pkg/tool/linux_ppc64le/compile: signal: killed

— watchflakes

[bcmills]

(attn @golang/ppc64)

[pmur]

Thanks. I suspect this was VM configuration issue which is hopefully resolved now.

Background, the VM was setup with too many vcpus (160), and only 30GB of RAM. At the time, OSU had some issues configuring the vcpu count. It's set appropriately now (24, as it needed to be a multiple of 8). Those unused cores may have caused excess RAM usage on the host.

[cherrymui]

Thanks @pmur ! Sounds like we can close this for now. We can reopen if this happens again.