Release v0.27.0

Added

• Added option --prometheus.health-check-request-timeout to set the timeout used for health-check requests to the prometheus endpoint, defaulting to 5s. (#268)

Release v0.26.0

Added

• Added a leader election feature for HA prometheus setup to reduce the amount of data
  transported on the wire. The leader sidecar will send data as normal, while other non-leaders
  sidecars will be in a stand-by mode. (#264, #263)
• Added sidecar.metrics.current to report the number of active metrics grouped by metric type. (#259)

Changed

• limited the number of failing metrics reported via metrics and logs. (#261)

Release v0.25.0

Changed

• Single metric metadata fetching happens through /metadata/, not /targets/medatada, which
  should prevent causing too much stress on Prometheus (less data to query/fetch). (#252)

Release 0.24.0

Performance improvements, improved throughput.

• Assemble multiple points per Metric, limit requests by size instead of points. (#237)
• Replace --prometheus.max-timeseries-per-request flag with --opentelemetry.max-bytes-per-request with 64kB default size (#237)
• Rename --prometheus.max-shards to --opentelemetry.max-shards (#245)
• Rename --prometheus.min-shards to --opentelemetry.min-shards (#245)
• Adds new --opentelemetry.queue-size setting. (#247)

Release v0.23.1

Changed

• Add sidecar.series.current to the periodic supervisor log. (#236)
• Fix issue w/ nextSegment being set incorrectly. (#242)

Release v0.23.0

Changed

• Do not require in-order writes, use random load balancing. (#198)
• New metrics sidecar.refs.collected and sidecar.refs.notfound count series references removed in garbage collection and not found during lookup. (#203)
• Added a loop at the start of Tail to wait in the event that prometheus is writing a new checkpoint. The max period for this loop is 5m0s (#205)
• Add proper support for Summary (#207)
• Keep existing shards during re-sharding events. (#221)
• Instead of skipping segments, restart the Prometheus reader to ensure we don't miss series references. (#227)
• Do not count Prometheus staleness markers as skipped points. (#231)

Release v0.22.0

• refs/tags/v0.22.0

Release v0.21.1

Removed

• Removed healthcheck metrics from telemetry traces

Release v0.21.0

Added

• New metric sidecar.points.skipped counts points that were not processed due to filters or cumulative resets. (#174)
• New metric sidecar.series.defined counts the number of series refs defined in the WAL. (#174)
• New metric sidecar.metadata.lookups counts the number of metadata lookups (with error=true/false). (#174)
• New metric sidecar.cumulative.missing_resets counts the number of points that were not processed due to cumulative resets. (#174)
• New metric sidecar.series.current reports the current number of series (with status=live/filtered/invalid). (#174)
• Added support for handling relabeling rules for instance label. (#175)

Changed

• Series cache remembers points that were filtered in order to correctly count points that are dropped. (#174)
• Metric sidecar.metadata.fetch.duration has new mode label for single and batch requests. (#174)
• Noisy logs are reduced to emitting once per minute. (#181)

Release v0.20.1

Changed

• The sidecar will no longer wait indefinitely when waiting for the initial scrape
  to complete, it will wait 60s longer than the longest missing interval. (#171)