[RELEASE] Scylla Monitoring Stack 3.9

[Amnon Heiman]

The Scylla team is pleased to announce the release of Scylla Monitoring Stack 3.9.0

Scylla Monitoring Stack is an open-source stack for monitoring Scylla Enterprise and Scylla Open Source, based on Prometheus and Grafana. Scylla Monitoring Stack 3.9 supports:

• Scylla Open Source versions 4.3, 4.4, the coming 4.5 release and the future 4.6

• Scylla Enterprise versions 2020.x and 2021.x

• Scylla Manager 2.3.x, 2.4.x, 2.5.x

This release brings new panels and graphs, new Advisor advice, as well as bug fixes and stability improvements.

Related Links

• Download Scylla Monitoring 3.9

• Scylla Monitoring Stack Docs

• Upgrade from Scylla Monitoring 3.x to 3.y

Versions updates Scylla Monitoring Stack 3.9

• Set Prometheus version to 2.29.1 #1515

• Set Grafana version to 8.1.1 #1515

• Update Loki to 2.3.0 #1515

• Update the Alertmanager version to 0.22.2 #1428

New in Scylla Advisor

• New Advisor feature: Warn when newer Scylla database or Scylla Monitoring versions are available (Learn more about Scylla Advisor here.) 

New Information in Scylla Dashboards

• New Keyspace dashboard

By default, Scylla does not report per-table metrics.

Keep in mind that enabling per-table metrics should be done with care and only if the number of tables is limited.

To enable the per-table metrics reporting you will need to start Scylla with --enable-keyspace-column-family-metrics true flag.

The keyspace dashboard shows the tables of each keyspace, for better performance, peek just the tables you are looking at from the dropdown menu at the top row.

The dashboard shows read and write per table and their latencies.

• Overview dashboard enhancements:

• Make the per-DC nodes table larger #1522

• Show that a repair is running in the node table #1420

• Updated look, with Grafana 8.x, new time series and pie chart panels.

Bug Fixes

• Inconsistent avg naming in overview screen #1393

• CURRENT_VERSION.sh is not reliable #1329

 

Operational Changes

• Pie chart plugin removed from the repository

• Prometheus single hyphen command line option is deprecated (though still supported). For example to set the retention period to 30 days use: start-all.sh -b "--storage.tsdb.retention.time=30d" instead of: start-all.sh -b "-storage.tsdb.retention.time=30d"

• Add a warning when configuring Prometheus destinations on a localhost.

• Remove the ad hoc filter from the dashboards filtering.

• Remove the per-version dashboard templates from the repository.

• Make the alertmanager persistent, by using attached host directory, similar to Prometheus 

• Add a per-container limits configuration (--limit flag)

• Add per-container value mapping (--volume)

• Add per-container command line parameter configuration (--param)