Skip to content

2.17.0 / 2020-03-24

Compare
Choose a tag to compare
@prombot prombot released this 24 Mar 20:34
v2.17.0
39e01b3

This release implements isolation in TSDB. API queries and recording rules are
guaranteed to only see full scrapes and full recording rules. This comes with a
certain overhead in resource usage. Depending on the situation, there might be
some increase in memory usage, CPU usage, or query latency.

  • [FEATURE] TSDB: Support isolation #6841
  • [ENHANCEMENT] PromQL: Allow more keywords as metric names #6933
  • [ENHANCEMENT] React UI: Add normalization of localhost URLs in targets page #6794
  • [ENHANCEMENT] Remote read: Read from remote storage concurrently #6770
  • [ENHANCEMENT] Rules: Mark deleted rule series as stale after a reload #6745
  • [ENHANCEMENT] Scrape: Log scrape append failures as debug rather than warn #6852
  • [ENHANCEMENT] TSDB: Improve query performance for queries that partially hit the head #6676
  • [ENHANCEMENT] Consul SD: Expose service health as meta label #5313
  • [ENHANCEMENT] EC2 SD: Expose EC2 instance lifecycle as meta label #6914
  • [ENHANCEMENT] Kubernetes SD: Expose service type as meta label for K8s service role #6684
  • [ENHANCEMENT] Kubernetes SD: Expose label_selector and field_selector #6807
  • [ENHANCEMENT] Openstack SD: Expose hypervisor id as meta label #6962
  • [BUGFIX] PromQL: Do not escape HTML-like chars in query log #6834 #6795
  • [BUGFIX] React UI: Fix data table matrix values #6896
  • [BUGFIX] React UI: Fix new targets page not loading when using non-ASCII characters #6892
  • [BUGFIX] Remote read: Fix duplication of metrics read from remote storage with external labels #6967 #7018
  • [BUGFIX] Remote write: Register WAL watcher and live reader metrics for all remotes, not just the first one #6998
  • [BUGFIX] Scrape: Prevent removal of metric names upon relabeling #6891
  • [BUGFIX] Scrape: Fix 'superfluous response.WriteHeader call' errors when scrape fails under some circonstances #6986
  • [BUGFIX] Scrape: Fix crash when reloads are separated by two scrape intervals #7011