By default, the Admin UI and each node of a CockroachDB cluster share anonymous usage details with Cockroach Labs. These details, which are completely scrubbed of identifiable information, greatly help us understand and improve how the system behaves in real-world scenarios.
This page summarizes the details that get shared, how to view the details yourself, and how to opt out of sharing.
What gets shared
When diagnostics reporting is on, each node of a CockroachDB cluster shares anonymized details on an hourly basis, including (but not limited to):
- Deployment and configuration characteristics, such as size of hardware, cluster settings that have been altered from defaults, number of replication zones configured, etc.
- Usage and cluster health details, such as crashes, unexpected errors, attempts to use unsupported features, types of queries run and their execution characteristics as well as types of schemas used, etc.
To view the full diagnostics details that a node reports to Cockroach Labs, use the
http://<node-address>:<http-port>/_status/diagnostics/local JSON endpoint.
In all cases, names and other string values are scrubbed and replaced with underscores. Also, the details that get shared may change over time, but as that happens, we will announce the changes in release notes.
Opt out of diagnostics reporting
At cluster initialization
To make sure that absolutely no diagnostic details are shared, you can set the environment variable
COCKROACH_SKIP_ENABLING_DIAGNOSTIC_REPORTING=true before starting the first node of the cluster. Note that this works only when set before starting the first node of the cluster. Once the cluster is running, you need to use the
SET CLUSTER SETTING method described below.
After cluster initialization
To stop sending diagnostic details to Cockroach Labs once a cluster is running, use the built-in SQL client to execute the following
SET CLUSTER SETTING statement, which switches the
diagnostics.reporting.enabled cluster setting to
> SET CLUSTER SETTING diagnostics.reporting.enabled = false;
This change will not be instantaneous, as it must be propagated to other nodes in the cluster.
Check the state of diagnostics reporting
> SHOW CLUSTER SETTING diagnostics.reporting.enabled;
diagnostics.reporting.enabled +-------------------------------+ false (1 row)
If the setting is
false, diagnostics reporting is off; if the setting is
true, diagnostics reporting is on.