Skip to content

OPERATIONS.md

Latest commit

 

History

History
42 lines (31 loc) · 1.65 KB

OPERATIONS.md

File metadata and controls

42 lines (31 loc) · 1.65 KB

Operations Runbook (dev)

This runbook addresses local tracking issues #26, #27 and #28 with deploy-time mitigations.

#26 Graph transient failures

  • Keep container restart policy enabled (restart: unless-stopped).
  • Prefer short poll cycles over very large mailboxes to reduce long-lived connections.
  • Collect and retain container logs; Graph failures are often intermittent network resets.
  • If using Kubernetes, use restartable workload types and alert on crash-loop frequency.

#27 since + watch performance

  • Split operation into two phases:
  • Backfill phase: finite since window, watch = false, explicit batch_size.
  • Steady-state phase: watch = true, smaller mailbox, shorter check interval.
  • Avoid first run against very large historical mailboxes with unbounded watch loops.

Recommended pattern:

  1. Backfill historical window (since = 7d, tune batch_size).
  2. Switch to steady state (since = 1d, watch = true).

#28 sink failures and exit code safety

  • Upstream may not always surface sink write failures as non-zero exit codes.
  • For one-shot/batch executions, wrap parsedmarc with scripts/run-parsedmarc-safe.sh.
  • The wrapper preserves parsedmarc output and forces non-zero exit for known sink error markers.
  • The wrapper now preserves original non-zero parsedmarc exit codes when no sink marker is present.

Example:

./scripts/run-parsedmarc-safe.sh -c /home/parsedmarc/ini/parsedmarc.ini

Notes:

  • The wrapper is a defensive mitigation, not a replacement for upstream fixes.
  • Keep mailbox delete/move behavior conservative until sink reliability is verified.

Validation:

./scripts/test-run-parsedmarc-safe.sh