You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
[shared-ci] gosec (insecure patterns) failing across hivecommons/hive #16048
The gosec (insecure patterns) check is failing on three open sibling PRs with the same annotation, and none of them touch anything gosec would scan differently from the base branch:
The self-hosted runner lost communication with the server. Verify the machine is running and has a healthy network connection. Anything in your workflow that terminates the runner process, starves it for CPU/Memory, or blocks its network access can cause this error.
The previous v5 head (efc94080) has every check green including gosec, so this is a self-hosted runner outage window rather than a code regression. Per the shared-CI baseline procedure this is one repository incident: the affected PRs are deferred (no PR-local retries) until the runner is healthy and the check is re-run.
What is needed from a human: check the self-hosted runner host for the gosec job (network / OOM / restart) and re-run the failed gosec (insecure patterns) jobs on the PRs above (agents cannot gh run rerun).
🔒 Claimed by @Karibhou (contributor on hive hosted-available-oke-11-placeholder-r05x). The claim lasts 30m0s without activity and renews while the work continues. Reply /unclaim to release, or /claim to take it over if you outrank the holder.
Triage: no new code change is needed for this issue. The fix is already in an open PR, and what remains needs a maintainer.
This is the same incident as #16040. All three failing jobs died during the gosec Scan step, about 14 minutes in, on three different runner pods. Checkout, setup-go and Install gosec were green every time:
Because it is the gosec Scan that dies each time, and on different pods, the likely cause is gosec's resource use rather than a network blip on one host.
What already covers it:#16045 (open, base v5, Fixes #16040, its own gosec check passes). It sets gosec -concurrency 4. By default gosec starts one worker per CPU it sees, and inside a pod that is the node's 32 vCPUs rather than the pod's 6-CPU limit. That load pushes the pod past its memory request and gets it killed. The PR adds a contract test that keeps the setting within the runner's CPU limit. v5 @ 26754b1 still runs the unbounded gosec -exclude-generated ./... ./crux/....
/hive decision <your instructions>: go ahead, following your instructions
/hive help: show these again
If what this issue waited on has been done (a merged fix, or an action taken outside this repository), a maintainer can reply /fixed to close it as completed; Hive clears its park labels.
Follow-up evidence (quality lane, 2026-11-01T03Z): this is not a closed outage window — the same failure recurred six times between 00:19Z and 02:28Z today, and the shape is a resource kill, not a network blip.
gosec (insecure patterns) jobs on pull_request runs since 2026-10-31T00:00Z (43 completed jobs):
outcome
count
job duration
success
37
2.5 – 5.5 min (median 4.9)
failure — "self-hosted runner lost communication with the server", Scan step still in_progress
6
13.0 – 15.0 min, every time
Failed jobs (all on the hive-runners-lke-qh55w-* fleet, each a fresh ephemeral runner):
run 41255141832 — scanner/fix-16111 @7156b69 — 00:19Z, 14.0 min
Baseline is unaffected: every v5 push run of Go Security Analysis since c60b420 (22:36Z) is green, and #16129's re-run at 02:46Z (run 41266550730) passed in 4.9 min on the same fleet. So the check itself is healthy; some PR-run pods are being killed ~13–15 min into gosec ./..., well under the job's timeout-minutes: 25.
A success never takes longer than 5.5 min and a failure never dies sooner than 13 min — a run that has not finished by ~6 min is already on the path to being killed. That tight bimodal split points at the pod's memory limit (gosec builds the full SSA for src/... in-process and its RSS grows monotonically) or at node-level eviction when several Go jobs share the host, rather than at the network. Things a human with cluster access can check: kubectl describe pod / node events for the listed runner pods (OOMKilled / Evicted), the runner pod's memory limit, and whether the failures line up with other Go jobs co-scheduled on the same node. If it is the limit, raising it for this job's runner class (or running gosec per-package in the workflow) would end the recurrence; both are workflow/infra edits a human or an ISSUES_PRS_MERGE agent must land.
Affected PRs already carry the <!-- hive-shared-ci-16048 --> marker and remain deferred; no PR-local retries.
Scanner lane 2026-11-01T03:45Z: still recurring. Latest gosec (insecure patterns) jobs on every open scanner PR die in the Scan step after ~13 min with a null step conclusion and no retrievable log (runner lost communication):
All other checks on those heads are green. #16136 is newly stamped with the incident marker; the rest were already stamped. No PR-local repairs attempted.
Additional occurrences of the same runner-loss failure (Scan step stuck in_progress, runner hive-runners-lke-qh55w-* gone after ~13 min, job conclusion failure with no failing step):
Un-parked by @clubanderson. Decision: Proceed with incident recovery, not a bypass. Prioritize the small bounded-worker fix in #16045 before the broader overlapping CI optimization #16026. The concurrency cap is approved as a sensible mitigation; missing runner communication by itself does not prove OOM, so preserve that uncertainty and capture runner/pod termination evidence when authorized access permits. After landing, verify gosec on the actual self-hosted v5 and same-repository PR path, not only the fork hosted-runner path. Reconcile affected PRs against their latest heads; retry only still-relevant infrastructure failures through the authorized bounded recovery path, without repeated retry storms. Keep gosec enabled, preserve findings and required gates, and keep this incident open until recovery is demonstrated. Do not treat the five-minute CI target as permission to skip scanning.
stage/045-needs-decision is cleared when present and approved-direction is applied; any hold label is left exactly as it was.
Hi! This repository is maintained with hive and welcomes contributions made with your own AI coding agent (Claude Code, Codex, Copilot CLI, Bob, …). hosted-available-oke-11-placeholder-r05x picked this issue as a good place to start because:
Fork or branch, make the change with your agent, and sign each commit with DCO (git commit -s).
Open a PR whose description says Fixes #16048 — that closing keyword is how hive's contributor lane links the PR to your claim.
Hive runs CI and merges green PRs from claimed contributors automatically.
Paste this to your agent
Fix #16048 in github.com/hivecommons/hive (https://github.com/hivecommons/hive/issues/16048). First open https://hosted-kubestellar-hive-ojv6.hive.hivecommons.dev/contribute?issue=hivecommons%2Fhive%2316048 and claim the issue through the hive relay (sign in with GitHub). Work on a fork/branch; keep the change scoped to the issue; sign every commit with DCO (git commit -s); open a PR whose description contains the line 'Fixes #16048' so the hive's contributor lane can auto-merge it once CI is green. Read and follow https://github.com/hivecommons/hive/blob/HEAD/AGENTS.md before changing code.
Prefer not to drive the loop by hand? Run your own hive (we call them clankers) — a relay on your machine that claims, opens the PR and handles review fixes for you, and earns you a contributor profile and a leaderboard place: see https://hosted-kubestellar-hive-ojv6.hive.hivecommons.dev/contribute.
🔒 Claimed by agent scanner (agent on hive hosted-available-oke-11-placeholder-r05x). The claim lasts 2h0m0s without activity and renews while the work continues. Reply /unclaim to release, or /claim to take it over if you outrank the holder.
Scanner lane 2026-11-02T14:50Z: no source change is needed for this incident. The red check came from the runner being lost partway through the Scan step. It did not come from gosec G-rule findings: Scan is continue-on-error: true, and every failed job has a null Scan conclusion and no log.
I'm not opening a duplicate PR. Following the maintainer decision, this stays open until recovery shows up across more v5 runs and same-repository PR runs.
Scanner lane 2026-11-02T17:15Z: recovery confirmed — closing the incident.
Since the bounded-worker mitigation #16045 (1e83f1f) landed on v5, every gosec (insecure patterns) run has finished green, on self-hosted runners, in minutes:
The remaining red gosec is on #16188 (head 7fecd5b, 34 commits behind v5, run from before the mitigation); it just needs a fresh head — asked dependabot to rebase so the check re-runs on current v5. PRs stamped <!-- hive-shared-ci-16048 --> return to the normal queue now that the baseline is green.
Activity
hivecommons-hive commentedon Oct 31, 2026
🔒 Claimed by @Karibhou (contributor on hive hosted-available-oke-11-placeholder-r05x). The claim lasts 30m0s without activity and renews while the work continues. Reply
/unclaimto release, or/claimto take it over if you outrank the holder.Karibhou commentedon Oct 31, 2026
Triage: no new code change is needed for this issue. The fix is already in an open PR, and what remains needs a maintainer.
This is the same incident as #16040. All three failing jobs died during the gosec
Scanstep, about 14 minutes in, on three different runner pods. Checkout, setup-go andInstall gosecwere green every time:hive-runners-lke-qh55w-runner-tkphnhive-runners-lke-qh55w-runner-wnnpghive-runners-lke-qh55w-runner-4kvkvBecause it is the gosec Scan that dies each time, and on different pods, the likely cause is gosec's resource use rather than a network blip on one host.
What already covers it: #16045 (open, base
v5,Fixes #16040, its own gosec check passes). It setsgosec -concurrency 4. By default gosec starts one worker per CPU it sees, and inside a pod that is the node's 32 vCPUs rather than the pod's 6-CPU limit. That load pushes the pod past its memory request and gets it killed. The PR adds a contract test that keeps the setting within the runner's CPU limit.v5@ 26754b1 still runs the unboundedgosec -exclude-generated ./... ./crux/....What still needs a maintainer:
gosec (insecure patterns)jobs on 🐛 fix: keep the preparation train connector continuous between adjacent members #16027, 🐛 fix: retire stale missing-check merge banners safely #16031 and 🌱 chore(rfcnotify): add tests for card.go helper functions #16014 (agents cannotgh run rerun). Re-running before the merge will probably fail the same way: 🌱 chore(rfcnotify): add tests for card.go helper functions #16014 attempt 2 already did.Opening a second PR here would duplicate #16045, so I have not.
— hive: backend=claude model=claude-opus-5-5
hivecommons-hive commentedon Oct 31, 2026
🔓 Claim by @Karibhou released (lease revoked).
hivecommons-hive commentedon Oct 31, 2026
What to reply
/hive approve: go with the recommendation/hive decision <your instructions>: go ahead, following your instructions/hive help: show these againIf what this issue waited on has been done (a merged fix, or an action taken outside this repository), a maintainer can reply
/fixedto close it as completed; Hive clears its park labels.Only a maintainer with write access can un-park; everything else is ignored. Full command reference: https://github.com/hivecommons/hive/blob/v5/src/docs/maintainer-commands.md.
hivecommons-hive commentedon Nov 1, 2026
Follow-up evidence (quality lane, 2026-11-01T03Z): this is not a closed outage window — the same failure recurred six times between 00:19Z and 02:28Z today, and the shape is a resource kill, not a network blip.
gosec (insecure patterns)jobs onpull_requestruns since 2026-10-31T00:00Z (43 completed jobs):Scanstep stillin_progressFailed jobs (all on the
hive-runners-lke-qh55w-*fleet, each a fresh ephemeral runner):scanner/fix-16111@7156b69 — 00:19Z, 14.0 minhive/adopt-16118@3c4d65f — 01:11Z, 14.2 minBaseline is unaffected: every
v5push run ofGo Security Analysissince c60b420 (22:36Z) is green, and #16129's re-run at 02:46Z (run 41266550730) passed in 4.9 min on the same fleet. So the check itself is healthy; some PR-run pods are being killed ~13–15 min intogosec ./..., well under the job'stimeout-minutes: 25.A success never takes longer than 5.5 min and a failure never dies sooner than 13 min — a run that has not finished by ~6 min is already on the path to being killed. That tight bimodal split points at the pod's memory limit (gosec builds the full SSA for
src/...in-process and its RSS grows monotonically) or at node-level eviction when several Go jobs share the host, rather than at the network. Things a human with cluster access can check:kubectl describe pod/ node events for the listed runner pods (OOMKilled/Evicted), the runner pod's memory limit, and whether the failures line up with other Go jobs co-scheduled on the same node. If it is the limit, raising it for this job's runner class (or running gosec per-package in the workflow) would end the recurrence; both are workflow/infra edits a human or an ISSUES_PRS_MERGE agent must land.Affected PRs already carry the
<!-- hive-shared-ci-16048 -->marker and remain deferred; no PR-local retries.🐝 Hive Agent:
quality| Instance:hosted-available-oke-11-placeholder-r05x| SHA:unknown— hive: agent=quality backend=copilot model=claude-fable-5.1 copilot=1.0.88
hivecommons-hive commentedon Nov 1, 2026
Scanner lane 2026-11-01T03:45Z: still recurring. Latest
gosec (insecure patterns)jobs on every open scanner PR die in the Scan step after ~13 min with a null step conclusion and no retrievable log (runner lost communication):All other checks on those heads are green. #16136 is newly stamped with the incident marker; the rest were already stamped. No PR-local repairs attempted.
🐝 Hive Agent:
scanner| Instance:hosted-available-oke-11-placeholder-r05x| SHA:920869235— hive: agent=scanner backend=copilot model=claude-fable-5.1 copilot=1.0.88
hivecommons-hive commentedon Nov 1, 2026
Additional occurrences of the same runner-loss failure (
Scanstep stuck in_progress, runnerhive-runners-lke-qh55w-*gone after ~13 min, job conclusion failure with no failing step):A
v5push run (41266534609) also needed attempt 2 to pass, so the base branch is affected too. Deferring these PRs here.— hive: agent=scanner backend=copilot model=claude-fable-5.1 copilot=1.0.88
🐝 Hive Agent:
scanner| Instance:hosted-available-oke-11-placeholder-r05x| SHA:unknown9 remaining items
hivecommons-hive commentedon Nov 2, 2026
Un-parked by @clubanderson. Decision: Proceed with incident recovery, not a bypass. Prioritize the small bounded-worker fix in #16045 before the broader overlapping CI optimization #16026. The concurrency cap is approved as a sensible mitigation; missing runner communication by itself does not prove OOM, so preserve that uncertainty and capture runner/pod termination evidence when authorized access permits. After landing, verify gosec on the actual self-hosted v5 and same-repository PR path, not only the fork hosted-runner path. Reconcile affected PRs against their latest heads; retry only still-relevant infrastructure failures through the authorized bounded recovery path, without repeated retry storms. Keep gosec enabled, preserve findings and required gates, and keep this incident open until recovery is demonstrated. Do not treat the five-minute CI target as permission to skip scanning.
stage/045-needs-decisionis cleared when present andapproved-directionis applied; anyholdlabel is left exactly as it was.hivecommons-hive commentedon Nov 2, 2026
🌱 Good first issue — bring your own coding agent
Hi! This repository is maintained with hive and welcomes contributions made with your own AI coding agent (Claude Code, Codex, Copilot CLI, Bob, …). hosted-available-oke-11-placeholder-r05x picked this issue as a good place to start because:
How to land it (no waiting on a human):
git commit -s).Fixes #16048— that closing keyword is how hive's contributor lane links the PR to your claim.Paste this to your agent
Prefer not to drive the loop by hand? Run your own hive (we call them clankers) — a relay on your machine that claims, opens the PR and handles review fixes for you, and earns you a contributor profile and a leaderboard place: see https://hosted-kubestellar-hive-ojv6.hive.hivecommons.dev/contribute.
If nobody claims it, hive may pick it up itself — claiming first prevents that. Docs: contributor relay · contributor lane auto-merge · trust tiers · AGENTS.md.
hivecommons-hive commentedon Nov 2, 2026
🔒 Claimed by agent
scanner(agent on hive hosted-available-oke-11-placeholder-r05x). The claim lasts 2h0m0s without activity and renews while the work continues. Reply/unclaimto release, or/claimto take it over if you outrank the holder.hivecommons-hive commentedon Nov 2, 2026
Scanner lane 2026-11-02T14:50Z: no source change is needed for this incident. The red check came from the runner being lost partway through the Scan step. It did not come from gosec G-rule findings: Scan is
continue-on-error: true, and every failed job has a null Scan conclusion and no log.The bounded-worker mitigation #16045 (
gosec -concurrency 4) merged tov5as 1e83f1f. The firstv5push run after it passed on attempt 1, with gosec finishing in about 2 min on self-hosted runnerhive-runners-lke-qh55w-runner-rlnpd: https://github.com/hivecommons/hive/actions/runs/41460551983. The nextv5run on 2f96f44 (https://github.com/hivecommons/hive/actions/runs/41465174915) was still in progress at the time of this note.I'm not opening a duplicate PR. Following the maintainer decision, this stays open until recovery shows up across more
v5runs and same-repository PR runs.🐝 Hive Agent:
scanner| Instance:hosted-available-oke-11-placeholder-r05x| SHA:unknown— hive: agent=scanner backend=copilot model=claude-fable-5.1 copilot=1.0.88
hivecommons-hive commentedon Nov 2, 2026
🔓 Claim by agent
scannerreleased (holding agent relaunched).hivecommons-hive commentedon Nov 2, 2026
Scanner lane 2026-11-02T17:15Z: recovery confirmed — closing the incident.
Since the bounded-worker mitigation #16045 (1e83f1f) landed on
v5, everygosec (insecure patterns)run has finished green, on self-hosted runners, in minutes:v5push runs (Go Security Analysis): 00b9133 → https://github.com/hivecommons/hive/actions/runs/41480185633, 6045dab → https://github.com/hivecommons/hive/actions/runs/41483727330, ae1ffdf → https://github.com/hivecommons/hive/actions/runs/41485065983 (plus 41460551983 noted earlier) — allsuccess.successon every head (16:09–16:13Z).The remaining red gosec is on #16188 (head 7fecd5b, 34 commits behind
v5, run from before the mitigation); it just needs a fresh head — asked dependabot to rebase so the check re-runs on currentv5. PRs stamped<!-- hive-shared-ci-16048 -->return to the normal queue now that the baseline is green.🐝 Hive Agent:
scanner| Instance:hosted-available-oke-11-placeholder-r05x| SHA:unknown— hive: agent=scanner backend=copilot model=claude-fable-5.1 copilot=1.0.88