ci: trade test parallelization for unconstrained gomaxprocs #12299
Conversation
| } | ||
| } | ||
|
|
||
| // Parallel runs t in parallel, unless CI is set to a true value. |
There was a problem hiding this comment.
Yes! I made it! You have no idea how hard Chrome had to work to let me add a comment here. After minutes of loading and many seconds to pop up this textbox after I clicked the line, my typing is limited to about 0.5 character / s.
Anyway... mind adding a comment here about why we disable parallel for CI? I assume the answer is "because we manually parallelize tests across multiple CI jobs" but it would be nice to leave a comment for future generations who find their way here (and perhaps a code change or move has obliterated this PR from git blame)
There was a problem hiding this comment.
fixed
And thank you for your sacrifice!
ci: trade test parallelization for unconstrained gomaxprocs
ci: trade test parallelization for unconstrained gomaxprocs
* Merge pull request #11889 from hashicorp/build-update-circle build: upgrade circleci configuration * Merge pull request #12299 from hashicorp/ci-parallel ci: trade test parallelization for unconstrained gomaxprocs * Merge pull request #12321 from hashicorp/ci-less-logging ci: limit gotestsum to circle ci * CI: build binaries for UI branches (#12594) Build binaries for every code change, not just backend code changes. This means that we'll have up-to-date compiled assets for every commit available in CircleCI artifacts. * Merge pull request #12736 from hashicorp/build-update-go-1.17.9 build: update golang to 1.17.9 * cgutil test: reserve only a single CPU for AddAlloc test Reserving the entire machine's worth of CPUs for the `AddAlloc` test triggers a condition where the empty shared CPUs defaults to the parent CPU set on some system configurations. This was done in `main` as part of cgroups v2 work, but we need to backport this to earlier branches in order to use the same machines across branches. Co-authored-by: Seth Hoenig <seth.a.hoenig@gmail.com> Co-authored-by: Tim Gross <tgross@hashicorp.com>
* Merge pull request #11889 from hashicorp/build-update-circle build: upgrade circleci configuration * Merge pull request #12299 from hashicorp/ci-parallel ci: trade test parallelization for unconstrained gomaxprocs * Merge pull request #12321 from hashicorp/ci-less-logging ci: limit gotestsum to circle ci * CI: build binaries for UI branches (#12594) Build binaries for every code change, not just backend code changes. This means that we'll have up-to-date compiled assets for every commit available in CircleCI artifacts. * cgutil test: reserve only a single CPU for AddAlloc test Reserving the entire machine's worth of CPUs for the `AddAlloc` test triggers a condition where the empty shared CPUs defaults to the parent CPU set on some system configurations. This was done in `main` as part of cgroups v2 work, but we need to backport this to earlier branches in order to use the same machines across branches. Co-authored-by: Seth Hoenig <seth.a.hoenig@gmail.com> Co-authored-by: Tim Gross <tgross@hashicorp.com>
|
I'm going to lock this pull request because it has been closed for 120 days ⏳. This helps our maintainers find and focus on the active contributions. |
This PR swaps
t.Parallelwithci.Parallel(t), which disables running in parallel in CI. The trade-off is we can now no longer restrictGOMAXPROCS=1. That plus upping the Circle CI go machine image tolarge, we should get CI runs in about 15 minutes.This is part 1 of $several for splitting up #12255; which aims to run CI on GHA.
Bad news: 419 files modified.
Good news: Only
config.yamlandslow.gocontain non-trivial changes.