Release v0.20.0

Changed:

• Uses Prometheus /api/v1/status/config endpoint to read the Prometheus config, to automatically determine the full set of scrape intervals. (#162)
• The startup timeout is raised to 10 minutes. (#166)

Removed:

• The --prometheus.scrape-interval option is ignored. (#162)

Release v0.19.0

Added

• Adding --healthcheck.threshold-ratio to support tuning the acceptable error ratio
  when exporting metrics to a backend. (#146)
• Print metadata from gRPC response trailers. (#151)
• Added sidecar.segment.skipped counter to keep track of the number of times an
  event has caused the WAL to be skipped. (#155)
• Parsing and reporting on dropped metric points due to validation errors
  using Lightstep's conventions. (#157)

Changed

• Fix metadata type conflict causing infinite loop due to change of instrument
  from histogram to another kind. (#151)
• Update Prometheus go.mod dependencies to match the 2.24.1 release. (#152)
• Update to OTel-Go 0.18. (#153)
• PrometheusReader handles truncated segment errors by raising an ErrSkipSegment which
  will trigger the tailer to skip to the next segment in process. (#155)
• Update to google.golang.org/protobuf v1.25.0, remove gogo dependency. (#156)

Removed

• Field corrupt-segment has been removed from the progress file as the state is
  no longer needed now that the PrometheusReader handles this case. (#155)

Release v0.18.3

Changed

• Fix issue that caused a segmentation failure on clean exit. (#143)
• Fix reset handling by checking against previous value instead of reset value. (#145)

Added

• Added a check for minimum version of prometheus on start, report an error and exit if prometheus running
  is less version 2.10.0. (#144)

Release v0.18.2

• Reduce the default value for max timeseries per request to 500. (#139)

Release v0.18.1

Changed

• set the default max shard value to 200

Release v0.18.0

Added

• Sidecar waits for the first scrapes to complete before entering its
  run state. (#134)
• New setting --prometheus.scrape-interval supports configuring scrape
  interval(s) to wait for at startup. (#134)

Changed

• The sidecar's WAL reader could get stuck in a restart loop in the event
  that the WAL's first segment after a checkpoint was truncated. The reader will
  now record the corrupt-segment in the progress log and skip the recorded
  segment on next restart (#136)

Removed

• The --startup.delay setting has been removed in favor of monitoring when
  Prometheus actually finishes its first scrapes. (#134)

Release v0.17.0

Added

• Automatically set (the same) service.instance.id for Destination/Diagnostics
  Resources. (#127)
• The sidecar's max timeseries per requests is now configurable
  via prometheus.max-timeseries-per-request. There is also a matching yaml configuration
  option: max_timeseries_per_request. (#128)
• The sidecar's max shards is now configurable via --prometheus.max-shards. There
  is also a matching yaml configuration option: max_shards. (#128)
• Adding metrics to capture WAL size and the current offset. (#130)

Changed

• The sidecar's WAL-reader addresses several race conditions by monitoring
  Prometheus for readiness and the current segment number during WAL segment
  transitions. (#118)
• The yaml section named "log_config" was inconsistent, has been renamed "log". ()
• Default values for max_shards and max_timeseries_per_request updated from 200 to 2000

Removed

• Remove the use_meta_labels parameter. (#125)

Release v0.16.0

Added

• The sidecar's health check period is now configurable via --healthcheck.period. The metrics reporting
  period is automatically configured to half of the healthcheck period, since one depends on the other. (#117)
• Adds support for gzip compression and none compression, because "" can't be configured via YAML. (#117)

Changed

• Improved liveness checking. The sidecar starts in a healthy state and if it passes its selftest it then begins
  judging its own health after 5 healthcheck periods. After liveness fails, the supervisor will very rapidly report
  a crash report and shutdown its span reporter to flush its diagnostics. There is a chance that k8s may kill the
  process before the crash report is sent. (#117)
• Numerous small consistency and style improvements for logs in general. (#117)

Release v0.15.0

In this release:

• disabled target refresh and caching functionality

Version 0.14.0: Observability improvements; snappy compression

• Timeouts and diagnostics for Prometheus API calls (#100)
• Snappy compression support enabled by default (#97)
• Supervisor will kill the sidecar when there are no successful writes after repeated healthchecks. (#101)
• Print the number of dropped series in the supervisor health report. (#102)