Release 4.8.1

Bug fixes

• start-all.sh --target-directory option have error in its documentation #2398
• Some panels in the OS dashboard do not respect the DC filter #2396
• Unknown Alternator OP - BatchGetItemSize #2393
• Compression-related panels are confusing #2392
• Using stack graph for active read is confusing #2389
• Row and Partitions insertions are measured in read/sec #2386
• The latencies Legend format are confusing #2383
• scylla_io_queue_flow_ratio graph is inconvenient #2382
• Add batch latency and batch size metrics to Alternator dashboard #2380
• Alternator OPs are not representative of real ops - in case of BatchGetItem and similar batch ops area/alternator #2374

Release 4.8.0

New In Release 4.8.0

• Support for Scylla Manager 3.3 #2339
• Make the Tablet section collapsible #2329
• Add panels for network compression #2325
• Add filters that limit the number of results per panel [breaking-changes] #2319
• Add a graph for scylla_io_queue_flow_ratio #2306
• Make the IO-group panel group by iogroup, stream #2305
• Tooltip now allows scrolling #2209
• Add metrics for RPC #2104
• Unify Scylla-Manager status and progress #2009
• Different aggregation functions for the latency metrics #1741

Bug Fixes

• Non-Paged CQL Reads Gauge isn't working. #2295
• " I/O Group All Queue consumption" dashboard use wrong type of graph. #2293
• Increase nodes table column width to display full ip by default #2302
• Fix panels description in the advanced dashboard #2290
• Full page screenshot is broken #2324
• Make genconfig support ipv6

Operational changes

• splitBrain alert support for a multi-cluster setup #2304
• Allow setting local network and docker_pram from env file #2035
• scylla_storage_proxy_coordinator_read_timeouts repeated twice in regexp. #2323
• make prometheus the default datasource #2268
• Deprecated level label #2322
• The os dashboard accepts multiple node_exporter jobs #2317
• Support Prometheus various scrap interval sampling [breaking-changes] #2345

Breaking Changes

• The dashboards now support longer Prometheus scrape intervals, which are configurable and passed as a parameter in the Grafana data source configuration.
• To better handle clusters with high core counts, the dashboards limit the number of series shown by default. You can change that limit from the drop-down menu at the top.

Release 4.7.2

Bug Fixes in Release 4.7.2

• Alternator, complete the move to summaries #2278
• Wrong port number for node manager agents in prometheus/prometheus.consul.yml.template when using Docker #2277

Release 4.7.1

Bug fixes in 4.7.1

• The bloom filter alert causes too many false positives #2263
• Non-token aware queries graph (and gauge) is broken. #2259
• Service level selection is not carried over between dashboards #2253
• Hints manager sent annotation, uses the wrong metric #2250
• Add cluster label to manager base metrics #2270

Release 4.7.0

New in Release 4.7.0

• Update alternator dashboard #2226
• Make the default dashboard refresh interval configurable #2220
• Show scylla_sstables_bloom_filter_memory_size on the detailed dashboard #2219
• Update Alternator latencies histogram and summaries #2214
• Combine the Advisor table with the alert table in the overview dashboard #2166
• Easier method to run multiple monitoring stacks side-by-side #2164
• Add ethtool metrics to Datadog integration #2163
• Add tablet metrics to the detailed dashboard #2119, #2111
• Add storage-related metrics #2044
• New alert - cluster in split-brain state #1677
• Enhanced experience with --archive command line flag #2158, #2177
• The explanation for the unified class group graph is not clear #2178

Bug fixes

• No closing parenthesis #2229
• The variable $sg is not defined. #2228
• Prometheus continues to trigger alerts for a node that has already been removed from scylla_servers.yml #2227
• read-timeouts in the overview dashboard are breaking when no cdc metrics are reported #2193
• Manager metrics are inconsistent #2191
• Version information is cut - although there's plenty of space available in the panel #2189
• Reads panel does not reflect shards #2171
• Overview page - no data [write latency, Read timeout by DC] #2162
• Manager memory metrics interfere with the OS ones #2198
• The actual interval for calculating metrics is greater than the one specified in evaluation_interval. #2087

operational chagnes

• start-all.sh optionally skip alertmanager #2239
• Allow an easy way to start Prometheus with protobuf support #2155
• Regex for empty string |$^ in dashboards #2192
• prometheus/prometheus.yml.template: set evaluation interval to 20s #2185
• Improved experience when working with Archive #2177
• start-all.sh: create a file with the parameters of the last run operation #2174
• remove the deprecated level label #2160
• Performance and security enhencements #2154
• Allow setting local network from env file #2035

scylla-monitoring-4.6.2

Release 4.6.2

Release 4.6.1

New in release 4.6.1

• Alert severity for repair and backup failures changed to warn #2151
• Update Grafana version to 10.2.3

Bug Fixes

• support custom port when using podman #2152

Release 4.6.0

New in Release 4.6.0

• Add scylla_io_queue_consumption plots #2088
• Create a metric that shows cache hit/miss rate per table #285
• Add a section that shows all scheduling groups on the same graph #2121
• Add logged/unlogged batches graphs #2081
• The 'Timeouts' item in the general monitoring dashboard has no description #2056
• Add a "CQL Connections (creation) rate" graph #2053
• Alert Manager Rule: add a too-many-files alert #2060

Bug Fixes

• By Instance, doesn't work on any of the dashboards #2138
• "Request Shed" is supposed to be in the "Coordinator" section #2124
• Request/Response payload sizes units are wrong #2083

Operational Change

• support IPV6 without specifying ports #2136

• Prometheus does not start with an external directory, and sudo #2134

• Experimental - Use compose to start the monitoring stack #2123

• scylla-overview: Do not rely on recording rules for picking the scheduling group #2120

• remove thrift from all calculations #2102

• Need to support node_exporter ports #2092

Release 4.5.1

Bug fixed in 4.5.1

• Grafana renderer is missing #2097

Alerts severities update

• AlertManager: TooManyFiles generates false alarms for a partition with the OS #2113
• NoCql alerts should be 'info' #2105
• InstanceDown alerts should be 'error' #2101

Release 4.5.0

New In Scylla Monitoring 4.5.0

• move to grafana 10.1.5 and prometheus 2.47.1 #2023
• Support multiple priority group in the Overview dashboard #2013
• Split the read-timeout into categories #2085
• Expose all columns of system.large_xx tables #1991
• Track UJ node progress #2076

Bug fixes

• Bogus value on the Repair Progress graph #2019
• Timeouts are not properly displayed #2085

Operation Changes

• support taking targets from directory #1995
• Monitoring stack doesn't respect manager_agent.prometheus config parameter #1992
• Support for Apple Silicon Valley Chips enhancement #1956
• Easier datadog integration #1998
• Remove reverse order read warnings #1498