Prometheus 2.8 offers block-bulks a leg-up


Monitoring system and time series database Prometheus 2.8 is now available, introducing experimental vertical query merging and vertical compacting as well as minor performance improvements.

The notion to add a way of querying and compacting time overlapping blocks first came up in summer 2017, but was only really pushed forward when issues in the prometheus-at-scale project Thanos kept coming up. Since the feature is still experimental, it needs to be enabled via the –storage.tsdb.allow-overlapping-blocks flag to help with things like bulk import of already compacted blocks and get more resilient compaction.

To steal less time from admins’ hands, the Prometheus developers did also improve the performance of queries and reduce the time alert managers are in flux once reloaded. The remote_write API now uses write ahead logging, while an upgrade to Bootstrap 4 brought some UI enhancements. Future releases will see a replacing of the /graph page, so just know that something is in the works if you think this doesn’t look quite right yet.

Speaking of UI, when using the status/targets site, once refreshed the page now remembers you clicked on unhealthy if you looked for those particular targets. This means you don’t have to go through the trouble of switching from all to unhealthy anymore after a reload. Other than that histogram_quantile now merges buckets with equivalent le values to help with OpenMetrics transitions and get rid of inconsistencies.


Bug fixes that found their way into the release include one that takes care of issues when sorting rule groups, and one that should reinstate the support of password_file and bearer_token_file in Kubernetes SD. A complete list of changes can be found in the project’s GitHub repository.

Prometheus is licensed under the Apache License 2.0 and was initially developed at SoundCloud in 2012. It is now a graduated project under the umbrella of the Cloud Native Computing Foundation, which it joined in 2016.

- Advertisement -