Severity-aware subscriptions
Page on critical. Email on warning. Drop info into a Slack channel.
Every subscription pins a trigger plus a severity tier, then routes to the channel that fits the level. PagerDuty channels accept a per-trigger severity_map (critical / error / warning / info) so task.failed pages oncall while agent.offline only nudges Slack. Combine with priority range, fnmatch task name pattern, and queue filters to narrow the firehose down to exactly what each tier should hear.
Ships with
- Six triggers covering tasks (failed / succeeded / retried / slow) and agents (online / offline)
- Severity tiers: critical / error / warning / info with sensible per-trigger defaults built in
- PagerDuty severity_default plus severity_map per trigger; one channel can route every level correctly
- Filter by priority_min / priority_max range so a low-priority background job never wakes anyone
- fnmatch patterns on task name (billing.*, *.deliver) and exact-match queue filters
- Project default subscriptions act as templates new members inherit; users override their own without affecting the team
Highlights
- Defaults are tuned so most operators only paste the integration key and go
- ReDoS-bounded matcher: complex severity_map regex skipped instead of pinning a worker
- Filters evaluated before dispatch, so a muted or out-of-range subscription costs nothing
Related