titan/pegasus
3
0
Fork 0
Code Issues 35 Pull Requests 1 Actions Packages Projects Releases Wiki Activity

[hermes] pegasus #303: transient_infra_failure #5

New Issue
Open
opened 2026-09-20 19:08:50 +00:00 by hermes-automation · 0 comments
hermes-automation commented 2026-09-20 19:08:50 +00:00
Member
Copy Link

Hermes auto-triage classified incident pegasus/303 as transient_infra_failure (confidence 0.95); first failed gate: scm_checkout.

Why a human is needed

Build pegasus #303 never reached its own pipeline: the very first step, fetching the repository from https://scm.bstein.dev/titan/pegasus.git to read the Jenkinsfile, failed with HTTP 502 from the Git host, and Jenkins aborted after exhausting its checkout retries. The 502 indicates the SCM service or its reverse proxy was momentarily unable to serve the request; nothing in the pegasus code, tests, or pipeline definition was exercised or is implicated. The fix is simply to re-run the build once scm.bstein.dev is responding; if 502s recur across jobs, the Gitea service and its ingress on the host side should be checked for restarts or resource pressure.

Ariadne did not authorize automated remediation: evidence_signature_missing.

Facts

  • jenkins — Build pegasus #303 finished FAILURE after 0.274 seconds (19:04:00.647 to 19:04:00.921 UTC), before any pipeline stage ran; first_failed_stage is null and failed_tests is empty. (https://ci.bstein.dev/job/pegasus/303/ (jenkins.result, jenkins.timestamps, jenkins.first_failed_stage, jenkins.failed_tests))
  • jenkins — The initial Jenkinsfile checkout failed: 'git fetch --tags --force --progress -- https://scm.bstein.dev/titan/pegasus.git' exited with status 128 and stderr 'fatal: unable to access https://scm.bstein.dev/titan/pegasus.git/: The requested URL returned error: 502'. (jenkins.console_failures[0] (line 13) and jenkins.console_tail)
  • jenkins — Jenkins retried the checkout until 'Maximum checkout retry attempts reached, aborting'; every attempt failed the same way and no build steps executed afterward. (jenkins.console_tail)
  • jenkins — The build was started by timer, not by a new commit push. (jenkins.console_tail, first line 'Started by timer')
  • opensearch — No correlated log records were available for the incident window; the OpenSearch query itself failed with 'Connection refused'. (log_evidence.error)

Evidence

The earliest failure region of the build console:

	at PluginClassLoader for git//hudson.plugins.git.GitSCM.retrieveChanges(GitSCM.java:1239)
	at PluginClassLoader for git//hudson.plugins.git.GitSCM._checkout(GitSCM.java:1310)
	at PluginClassLoader for git//hudson.plugins.git.GitSCM.checkout(GitSCM.java:1277)
	at PluginClassLoader for workflow-scm-step//org.jenkinsci.plugins.workflow.steps.scm.SCMStep.checkout(SCMStep.java:136)
	at PluginClassLoader for workflow-cps//org.jenkinsci.plugins.workflow.cps.CpsScmFlowDefinition.create(CpsScmFlowDefinition.java:168)
	at PluginClassLoader for workflow-cps//org.jenkinsci.plugins.workflow.cps.CpsScmFlowDefinition.create(CpsScmFlowDefinition.java:70)
	at PluginClassLoader for workflow-job//org.jenkinsci.plugins.workflow.job.WorkflowRun.run(WorkflowRun.java:317)
	at hudson.model.ResourceController.execute(ResourceController.java:101)
	at hudson.model.Executor.run(Executor.java:460)
Caused by: hudson.plugins.git.GitException: Command "git fetch --tags --force --progress -- https://scm.bstein.dev/titan/pegasus.git +refs/heads/*:refs/remotes/origin/*" returned status code 128:
stdout: 
stderr: fatal: unable to access 'https://scm.bstein.dev/titan/pegasus.git/': The requested URL returned error: 5
... (truncated; the full log is in the linked build)

Inferences

  • The Git host at scm.bstein.dev (or the reverse proxy in front of it) returned HTTP 502 for the fetch, meaning the proxy could not reach the Gitea backend at that moment; the failure occurred entirely outside the pegasus repository's code, tests, or Jenkinsfile.
  • Because the build was timer-triggered and never checked out source, there is no repository change to fix; re-running the same build after the SCM host is reachable again should succeed.
  • The empty OpenSearch result reflects a failure to query the log store during evidence collection, not an absence of activity, so it neither supports nor contradicts the Jenkins evidence.

Links

  • Failed build: https://ci.bstein.dev/job/pegasus/303/
  • Full evidence bundle and audit trail live in Ariadne at /api/admin/audit/events, event types hermes_autotriage_incident and hermes_autotriage_diagnosis.

Filed automatically by Ariadne from a Hermes Agent diagnosis (run run_1d499e6f63b945b5a8408de445bc8e99). Hermes has no write access to this repository; no files or infrastructure were changed.

Hermes auto-triage classified incident `pegasus/303` as **transient_infra_failure** (confidence 0.95); first failed gate: `scm_checkout`. ## Why a human is needed Build pegasus #303 never reached its own pipeline: the very first step, fetching the repository from https://scm.bstein.dev/titan/pegasus.git to read the Jenkinsfile, failed with HTTP 502 from the Git host, and Jenkins aborted after exhausting its checkout retries. The 502 indicates the SCM service or its reverse proxy was momentarily unable to serve the request; nothing in the pegasus code, tests, or pipeline definition was exercised or is implicated. The fix is simply to re-run the build once scm.bstein.dev is responding; if 502s recur across jobs, the Gitea service and its ingress on the host side should be checked for restarts or resource pressure. Ariadne did not authorize automated remediation: `evidence_signature_missing`. ## Facts - **jenkins** — Build pegasus #303 finished FAILURE after 0.274 seconds (19:04:00.647 to 19:04:00.921 UTC), before any pipeline stage ran; first_failed_stage is null and failed_tests is empty. (`https://ci.bstein.dev/job/pegasus/303/ (jenkins.result, jenkins.timestamps, jenkins.first_failed_stage, jenkins.failed_tests)`) - **jenkins** — The initial Jenkinsfile checkout failed: 'git fetch --tags --force --progress -- https://scm.bstein.dev/titan/pegasus.git' exited with status 128 and stderr 'fatal: unable to access https://scm.bstein.dev/titan/pegasus.git/: The requested URL returned error: 502'. (`jenkins.console_failures[0] (line 13) and jenkins.console_tail`) - **jenkins** — Jenkins retried the checkout until 'Maximum checkout retry attempts reached, aborting'; every attempt failed the same way and no build steps executed afterward. (`jenkins.console_tail`) - **jenkins** — The build was started by timer, not by a new commit push. (`jenkins.console_tail, first line 'Started by timer'`) - **opensearch** — No correlated log records were available for the incident window; the OpenSearch query itself failed with 'Connection refused'. (`log_evidence.error`) ## Evidence The earliest failure region of the build console: ``` at PluginClassLoader for git//hudson.plugins.git.GitSCM.retrieveChanges(GitSCM.java:1239) at PluginClassLoader for git//hudson.plugins.git.GitSCM._checkout(GitSCM.java:1310) at PluginClassLoader for git//hudson.plugins.git.GitSCM.checkout(GitSCM.java:1277) at PluginClassLoader for workflow-scm-step//org.jenkinsci.plugins.workflow.steps.scm.SCMStep.checkout(SCMStep.java:136) at PluginClassLoader for workflow-cps//org.jenkinsci.plugins.workflow.cps.CpsScmFlowDefinition.create(CpsScmFlowDefinition.java:168) at PluginClassLoader for workflow-cps//org.jenkinsci.plugins.workflow.cps.CpsScmFlowDefinition.create(CpsScmFlowDefinition.java:70) at PluginClassLoader for workflow-job//org.jenkinsci.plugins.workflow.job.WorkflowRun.run(WorkflowRun.java:317) at hudson.model.ResourceController.execute(ResourceController.java:101) at hudson.model.Executor.run(Executor.java:460) Caused by: hudson.plugins.git.GitException: Command "git fetch --tags --force --progress -- https://scm.bstein.dev/titan/pegasus.git +refs/heads/*:refs/remotes/origin/*" returned status code 128: stdout: stderr: fatal: unable to access 'https://scm.bstein.dev/titan/pegasus.git/': The requested URL returned error: 5 ... (truncated; the full log is in the linked build) ``` ## Inferences - The Git host at scm.bstein.dev (or the reverse proxy in front of it) returned HTTP 502 for the fetch, meaning the proxy could not reach the Gitea backend at that moment; the failure occurred entirely outside the pegasus repository's code, tests, or Jenkinsfile. - Because the build was timer-triggered and never checked out source, there is no repository change to fix; re-running the same build after the SCM host is reachable again should succeed. - The empty OpenSearch result reflects a failure to query the log store during evidence collection, not an absence of activity, so it neither supports nor contradicts the Jenkins evidence. ## Links - Failed build: https://ci.bstein.dev/job/pegasus/303/ - Full evidence bundle and audit trail live in Ariadne at `/api/admin/audit/events`, event types `hermes_autotriage_incident` and `hermes_autotriage_diagnosis`. Filed automatically by Ariadne from a Hermes Agent diagnosis (run [run_1d499e6f63b945b5a8408de445bc8e99](https://triage.bstein.dev/chat?resume=run_1d499e6f63b945b5a8408de445bc8e99)). Hermes has no write access to this repository; no files or infrastructure were changed. <!-- hermes-triage job=pegasus classification=transient_infra_failure incident=pegasus/303 -->
Sign in to join this conversation.
No Branch/Tag Specified
Branches Tags
Labels
No items
No Label
Milestone
No items
No Milestone
Projects
Clear projects
No project
Assignees
Clear assignees
atlas-reconciler bstein hermes-automation
No Assignees
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: titan/pegasus#5
No description provided.
Delete Branch "%!s()"

Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?