v5.1
StackState.comDownloadSupportExplore playground

Data retention

StackState Self-hosted v5.1.x

Overview

StackState imposes data retention limits to save storage space and improve performance. You can configure the data retention period to balance the amount of data stored with StackState performance and data availability.

Retention of topology graph data

By default, topology graph data will be retained for 8 days. This works in a way that the latest state of topology graph will always be retained; only history older than 8 days will be removed. You can check and alter the configured retention period this using the StackState CLI.

From StackState v5.0, the old sts CLI has been renamed to stac and there is a new sts CLI. The command(s) provided here are for use with the new sts CLI.

➡️ Check which version of the sts CLI you are running

$ sts graph retention

In some cases, it may be useful to keep historical data for more than eight days.

From StackState v5.0, the old sts CLI has been renamed to stac and there is a new sts CLI. The command(s) provided here are for use with the new sts CLI.

➡️ Check which version of the sts CLI you are running

$ sts graph retention --set 10d

(note that the duration can be specified as a duration string)

Note that by adding more time to the data retention period, the amount of data stored is also going to grow and need more storage space. This may also affect the performance of the Views.

After the new retention window is applied, you can schedule a new removal with this command:

From StackState v5.0, the old sts CLI has been renamed to stac and there is a new sts CLI. The command(s) provided here are for use with the new sts CLI.

➡️ Check which version of the sts CLI you are running

$ sts graph retenstion --set 10d --schedule-removal

Command not currently available in the new sts CLI. Use the stac CLI.

After changing the retention period to a smaller window, you may end up with some data that's already expired and will wait there until the next scheduled cleanup. To schedule an additional removal of expired data, use the following command:

Note that this may take some time to have an effect.

From StackState v5.0, the old sts CLI has been renamed to stac and there is a new sts CLI. The command(s) provided here are for use with the new sts CLI.

➡️ Check which version of the sts CLI you are running

$ sts graph delete-expired-data

However, if you would like to perform data deletion without having to wait for an additional scheduled cleanup, you can use --immediately argument:

From StackState v5.0, the old sts CLI has been renamed to stac and there is a new sts CLI. The command(s) provided here are for use with the new sts CLI.

➡️ Check which version of the sts CLI you are running

$ sts graph delete-expired-data --immediate

Retention of events, metrics and traces

StackState data store

If you are using the event/metrics/traces store provided with StackState, your data will by default be retained for 30 days. In most cases, the default settings will be sufficient to store all indices for this amount of time.

Configure disk space for Elasticsearch

In some circumstances it may be necessary to adjust the disk space available to Elasticsearch and how it's allocated to each index group, for example if you anticipate a lot of data to arrive for a specific index.

The settings can be adjusted by using environment variables to override the default configuration of the parameters described below.

Note that elasticsearchDiskSpaceMB will scale automatically based on the disk space available to Elasticsearch in Kubernetes.

ParameterDefaultDescription

elasticsearchDiskSpaceMB

400000

The total disk space assigned to Elasticsearch in MB. The default setting is the recommended disk space for a StackState production setup (400GB).

splittingStrategy

"days"

The frequency of creating new indices. Can be one of "none", "hours", "days", "months" or "years". If "none" is specified, only one index will be used.

maxIndicesRetained

30

The number of indices that will be retained in each index group. Together with the splittingStrategy governs how long historical data will be kept in Elasticsearch.

diskSpaceWeight

Varies per index group

Defines the share of disk space an index will get based on the total elasticsearchDiskSpaceMB. If set to 0 then no disk space will be allocated to the index. See the disk space weight examples below.

replicas

Linux: 0 Kubernetes: 1

The number of nodes that a single piece of data should be available on. Use for redundancy/high availability when more than one Elasticsearch node is available.

stackstate {
  ...

  // Total size of disk assigned to Elasticsearch in MB
  elasticsearchDiskSpaceMB = 400000

  ...

  // For each index group:
  // kafkaMetricsToES - the sts_metrics index
  // kafkaMultiMetricsToES - the sts_multi_metrics index
  // kafkaGenericEventsToES - the sts_generic_events index
  // kafkaTopologyEventsToES - the sts_topology_events index
  // kafkaStateEventsToES - the sts_state_events index
  // kafkaStsEventsToES - the sts_events index
  // kafkaTraceToES - the sts_trace_events index

  kafkaMultiMetricsToES {
    ...
    elasticsearch {
      index {
        splittingStrategy = "days"
        maxIndicesRetained = 30
        refreshInterval = 1s
        replicas = 0 // Default setup is single node
        diskSpaceWeight = 1
      }
    ...
    }
  }
  ...
}

Disk space weight examples

Use the diskSpaceWeight configuration parameter to adjust how available disk space is allocated across Elasticsearch index groups. This is helpful if, for example, you expect a lot of data to arrive in a single index. Below are some examples of disk space weight configuration.

Note that increasing the total limit or the diskSpaceWeight will increase the amount of data that can be stored in each index. If the total value of metrics received is too high, it could affect telemetry stream performance due to increased metrics processing time.

Allocate no disk space to an index group Setting diskSpaceWeight to 0 will result in no disk space being allocated to an index group. For example, if you aren't going to use traces, then you can stop reserving disk space for this index group and make it available to other index groups with the setting:

 kafkaTraceToES.elasticsearch.index.diskSpaceWeight = 0

Distribute disk space unevenly across index groups The available disk space (the configured elasticsearchDiskSpaceMB) will be allocated to index groups proportionally based on their configured diskSpaceWeight. Disk space will be allocated to each index group according to the formula below, this will then be shared between the indices in the index group:

# Total disk space allocated to an index group
index_group_disk_space = (elasticsearchdiskSpaceMB * diskSpaceWeight / sum(diskSpaceWeights)

For example, with elasticsearchDiskSpaceMB = 300000, disk space would be allocated to the index groups and indexes be as follows:

ParameterIndex group disk space

kafkaMetricsToES.elasticsearch.index { diskSpaceWeight = 0 maxIndicesRetained = 20 }

0MB

kafkaMultiMetricsToES.elasticsearch.index { diskSpaceWeight = 1 maxIndicesRetained = 20 }

20,000MB or 300,000*1/15

kafkaGenericEventsToES.elasticsearch.index{ diskSpaceWeight = 2 maxIndicesRetained = 20 }

40,000MB or 300,000*2/15

kafkaTopologyEventsToES.elasticsearch.index{ diskSpaceWeight = 3 maxIndicesRetained = 20 }

60,000MB or 300,000*3/15

kafkaStateEventsToES.elasticsearch.index{ diskSpaceWeight = 4 maxIndicesRetained = 20 }

80,000MB or 300,000*4/15

kafkaStsEventsToES.elasticsearch.index{ diskSpaceWeight = 5 maxIndicesRetained = 20 }

100000MB or 300,000*5/15

kafkaTraceToES.elasticsearch.index{ diskSpaceWeight = 0 maxIndicesRetained = 20 }

0MB

External data store

If you have configured your own data source to be accessed by StackState, the retention policy is determined by the metric/event store that you have connected.

PreviousConfiguration backupNextClear stored data

Last updated

Logo

Legal notices

PrivacyCookiesResponsible disclosureSOC 2/SOC 3