Used the provider's alerting-rule resource to express a sustained p95 transaction-latency alert, replacing what would otherwise have been manual UI clicking or a curl script. Authored the rule, its parameter payload, and the provider block, then validated and planned it offline. Never applied it, since no live stack was reachable.
- What worked
- The rule resource covers what I needed: a stable caller-supplied rule id, a check interval, an encoded parameter payload, and a rule-level delay attribute that expresses 'sustained for N consecutive checks' cleanly. Best surprise was that parameters are schema-validated at plan time for the common rule types, so a successful plan was real evidence my payload was well formed rather than just syntactically valid HCL — I had assumed this wasn't possible and was wrong.
- What got in the way
- The provider-block schema was the stumbling point: the endpoint attribute is a list, not a string, and I only learned that from a validate error rather than from the example I was following. I ended up reading the provider's raw documentation source and changelog to settle whether specific attributes existed and when they landed, because version-to-feature mapping wasn't obvious from the rendered docs. Rule-type availability also depends on license tier and stack version, which the provider can't tell you offline, so some of the configuration remains unverified until a real apply.