On a high-level this release:
- Improves ingest performance considerably, by at least 2x
- Improves memory usage
- Decreases CPU usage
- Handles ingest of many metrics better
- Fixes a bug with tables that are compressed and then backfilled
Changelog
b6498d0 Add a share lock for the series parent table
0e8577c Add basic telemetry
bb7617b Add option to output postgres log, and clean docker image when done
2286518 Create an insert ctx to reuse memory across requests
093802c Decompress automatically if trying to insert to a compressed chunk
9650f91 Fix bug in extension selector
232756e Fix log.go always being the caller in the logs
44f1597 Fix race conditions in elector and advisory lock
3bd7ebd Fork prometheus prompb inside our repo
09e08a4 Get rid of time-only index
a53c5a3 Increase number of database connections in the pool.
066695e Log a version and latest commit hash on startup
a565b92 Modify the protobufs to reuse allocated memory
e76a319 Optimize series inserts
6e75176 Prepare for the 0.1.0-alpha.4 release
461d52f Put in replacement rule for pgconn
e70f584 Reduce memory usage on ingest
f7b5fc0 Refactor Labels creation
4a7b4e4 Set version to last tag in Docker build
9a63194 Split metric creation into sync and async parts
eeb3e23 Switch to dedicated CopyFrom workers
d34591f Switch to per-metric inserter
a0cf084 Use the InsertCtx for Labels creation
Docker images
docker pull timescale/timescale-prometheus:0.1.0-alpha.4docker pull timescale/timescale-prometheus:0.1docker pull timescale/timescale-prometheus:latest
This is an alpha release, so there are NO upgrade guarantees.