There's been a number of performance improvements. WAL reading is now about four times faster, bringing faster startup times. The amount of RAM used by compaction, indexes, in-memory series, and low ingestion rate servers have all been notably reduced. Remote write performance has been improved, and you can now specify a minimum number of shards. Remote read is now resilient to failures of backends.
In terms of service discovery Azure, EC2 and Kubernetes got new metadata. There are some new metrics for Prometheus itself for the TSDB, templates, and rules. Relatedly, the
promtool update has been removed, so if you want to convert 1.x recording rules to the 2.x format you'll want to use a promtool from an earlier version. Prometheus now identifies itself in the user agent of outbound HTTP requests, and the console tab in the UI supports choosing the evaluation time of the instant query.
There’s also a number of improvements, bug fixes and other changes, see the release notes for more detail.
Wondering how to upgrade? Contact us.