The sidecar component of Thanos gets deployed along with a Prometheus instance. This allows sidecar to optionally upload metrics to object storage and allow Queriers to query Prometheus data with common, efficient StoreAPI.
NOTE: This still does NOT mean that Prometheus can be fully stateless, because if it crashes and restarts you will lose ~2 hours of metrics, so persistent disk for Prometheus is highly recommended. The closest to stateless you can get is using remote write (which Thanos experimentally supports, see this. Remote write has other risks and consequences, and still if crashed you loose in positive case seconds of metrics data, so persistent disk is recommended in all cases.
Prometheus servers connected to the Thanos cluster via the sidecar are subject to a few limitations and recommendations for safe operations:
(!) The Prometheus
external_labels section of the Prometheus configuration file has unique labels in the overall Thanos system. Those external labels will be used by the sidecar and then Thanos in many places:
--web.enable-lifecycle flag is enabled if you want to use sidecar reloading features (
If you choose to use the sidecar to also upload to object storage:
--storage.tsdb.max-block-durationmust be set to equal values to disable local compaction on order to use Thanos sidecar upload, otherwise leave local compaction on if sidecar just exposes StoreAPI and your retention is normal. The default of
2his recommended. Mentioned parameters set to equal values disable the internal Prometheus compaction, which is needed to avoid the uploaded data corruption when Thanos compactor does its job, this is critical for data consistency and should not be ignored if you plan to use Thanos compactor. Even though you set mentioned parameters equal, you might observe Prometheus internal metric
prometheus_tsdb_compactions_totalbeing incremented, don’t be confused by that: Prometheus writes initial head block to filesytem via its internal compaction mechanism, but if you have followed recommendations - data won’t be modified by Prometheus before the sidecar uploads it. Thanos sidecar will also check sanity of the flags set to Prometheus on the startup and log errors or warning if they have been configured improperly (#838).
Thanos can watch changes in Prometheus configuration and refresh Prometheus configuration if
You can configure watching for changes in directory via
Thanos sidecar can watch
--reloader.config-file=CONFIG_FILE configuration file, evaluate environment variables found in there, and produce generated config in
$ prometheus \ --storage.tsdb.max-block-duration=2h \ --storage.tsdb.min-block-duration=2h \ --web.enable-lifecycle
$ thanos sidecar \ --tsdb.path "/path/to/prometheus/data/dir" \ --prometheus.url "http://localhost:9090" \ --objstore.config-file "bucket.yml"
The example content of
type: GCS config: bucket: example-bucket
If you want to migrate from a pure Prometheus setup to Thanos and have to keep the historical data, you can use the flag
--shipper.upload-compacted. This will also upload blocks that were compacted by Prometheus.
To use this, the Prometheus compaction needs to be disabled. This can be done by setting the following flags for Prometheus:
usage: thanos sidecar [<flags>] sidecar for Prometheus server Flags: -h, --help Show context-sensitive help (also try --help-long and --help-man). --version Show application version. --log.level=info Log filtering level. --log.format=logfmt Log format to use. --tracing.config-file=<file-path> Path to YAML file with tracing configuration. See format details: https://thanos.io/tracing.md/#configuration --tracing.config=<content> Alternative to 'tracing.config-file' flag (lower priority). Content of YAML file with tracing configuration. See format details: https://thanos.io/tracing.md/#configuration --http-address="0.0.0.0:10902" Listen host:port for HTTP endpoints. --grpc-address="0.0.0.0:10901" Listen ip:port address for gRPC endpoints (StoreAPI). Make sure this address is routable from other components. --grpc-server-tls-cert="" TLS Certificate for gRPC server, leave blank to disable TLS --grpc-server-tls-key="" TLS Key for the gRPC server, leave blank to disable TLS --grpc-server-tls-client-ca="" TLS CA to verify clients against. If no client CA is specified, there is no client verification on server side. (tls.NoClientCert) --prometheus.url=http://localhost:9090 URL at which to reach Prometheus's API. For better performance use local network. --tsdb.path="./data" Data directory of TSDB. --reloader.config-file="" Config file watched by the reloader. --reloader.config-envsubst-file="" Output file for environment variable substituted config file. --reloader.rule-dir=RELOADER.RULE-DIR ... Rule directories for the reloader to refresh (repeated field). --objstore.config-file=<file-path> Path to YAML file that contains object store configuration. See format details: https://thanos.io/storage.md/#configuration --objstore.config=<content> Alternative to 'objstore.config-file' flag (lower priority). Content of YAML file that contains object store configuration. See format details: https://thanos.io/storage.md/#configuration --min-time=0000-01-01T00:00:00Z Start of time range limit to serve. Thanos sidecar will serve only metrics, which happened later than this value. Option can be a constant time in RFC3339 format or time duration relative to current time, such as -1d or 2h45m. Valid duration units are ms, s, m, h, d, w, y.