repaird

README

Repaird

This small deamon detects and fixes problems on each host. First use-case is fixing Ceph mounts on a host by detecting that they are stale.

Design

There is a detecting phase, if any problems are detected the action from is performed. Each distinct problem should live in its own *.go file in the repair/ subdirectory. All detectors are run serially, although this might change, if it proofs to become a problem.

• No logging should be done from these functions.
• Every detector will be run at rougly 1m interval.
• A maximum of 3 repairs will be attempted in 12m. If this maximum is reached, it will sleep for an hour.

Metrics

Per named problem:

• repair_attempted_count_total{name="<name>"} -- found something to repair for 'name'
• repair_failed_count_total{name="<name>"} -- repair failed for 'name'