01
- Rescue
- Instances with write errors, overlapping tasks, or a disk that just filled. We stabilize ingest first and write the incident up afterward.
02
- Cardinality review
- Tag design audited against the real queries, offenders found, and a plan to move them to fields or a different storage pattern without losing history.
03
- Retention and downsampling
- Retention policies, task schedules and bucket design set against what the dashboards actually ask for, so disk usage and query latency both come down.
04
- Version upgrades and migrations
- Moving between 1.x and 2.x, or planning the jump to 3.x, with parallel instances and a shadow query stage so Flux and InfluxQL both keep working.
05
- Telegraf and ingest
- Buffer sizes, flush intervals, outputs and backpressure handled so that write rejections become visible alerts rather than silent gaps.
06
- Managed Tech
- Cardinality alarms, task health, backup verification and on-call engineers for the night the TSM compaction decides to misbehave.