Run the E2E matrix on the 8-core runner - #271
Open
vitramir wants to merge 3 commits into
Open
Conversation
ubuntu-latest gives 4 vCPU and provision-vm hands all four to the VM, so qemu, chromium and the suite contend for the same cores. Lease renewals then stall past the 5 minute TTL, ziti GCs the identity, and tracing terminates mid-run -- the two tracing tests fail on a missing llm.call span with the workload otherwise healthy. cpus is passed here rather than raised in provision-vm's default: the action is consumed as @main by sixteen repos still on ubuntu-latest, where a 6 CPU default would over-commit a 4 CPU runner.
The runner is 8 cores / 32 GB, so the 12GiB default left most of the host's memory unused.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
ubuntu-latestgives 4 vCPU / 16 GB, andprovision-vmhands all four cores to the lima VM. qemu, the in-cluster suite and the host tooling then contend for the same four cores, with no headroom.That starvation is what the
services (orchestrator)job has been failing on. It reproduced twice in a row, both times on the two tracing tests:The workload itself was healthy every time (all pod conditions
True,ziti-enrollexit 0) and all 16 upstream LLM calls returned 200 — the spans simply never landed. When renewals stall past the 5 minute lease TTL, ziti GCs the identity and the holder terminates by design;changes/2026-07-23-service-identity-loss-fail-fast.mdrecords the same failure from host suspend, and CPU starvation produces the identical stall. On the VM the tracing pod has 166 restarts, every one of themziti identity lost.Both suites pass locally against the same build, where the VM is not fighting anything for cores.
cpusis passed from the workflow rather than raised in the action's default:provision-vmis consumed as@mainby sixteen repos still onubuntu-latest, and a 6 CPU default would over-commit a 4 CPU runner in all of them.Memory stays at the 12GiB default until we can see what the runner actually has.