Release v19.26.0

Added

• A chronicle-observer-quarantined counter records each time an observer is quarantined, so you can alert on observer quarantine as well as on partitions that ran out of retries. (#4427)

Changed

• The observer failure counters (chronicle-observer-partitions-failed, chronicle-observer-partition-retry-attempts, chronicle-observer-partitions-quarantined) no longer carry a partition tag. That tag held the event source id, which could identify a person and created one series per failing partition. They now have one series per observer, tagged by event store, namespace, observer and event sequence.
• Their exported names no longer carry their description as a unit suffix: in Prometheus they are chronicle_observer_partitions_failed_total, chronicle_observer_partition_retry_attempts_total and chronicle_observer_partitions_quarantined_total, and the description is now the metric's help text. Update dashboards and alert rules that use the old suffixed names; the rules in Get alerted when observers stop processing match both.
• The counters' descriptions now say what they count: chronicle-observer-partitions-failed counts every failed attempt, including each retry, not failed partitions.