Monitoring and Alerting

On this page

In addition to CockroachDB's built-in safeguards against failure, it is critical to actively monitor the overall health and performance of a cluster running in production and to create alerting rules that promptly send notifications when there are events that require investigation or intervention.

This page describes the monitoring and observability tools that are built into CockroachDB Self-Hosted and shows how to collect your cluster's metrics using external tools like Prometheus's AlertManager for event-based alerting. For more details, refer to:

• Monitor CockroachDB with Prometheus
• Third-party Monitoring Tools

To export metrics from a CockroachDB Cloud cluster, refer to Export Metrics From a CockroachDB Standard Cluster or Export Metrics From a CockroachDB Advanced Cluster instead of this page.

Built-in monitoring tools

CockroachDB includes several tools to help you monitor your cluster's workloads and performance.

DB Console

The DB Console collects time-series cluster metrics and displays basic information about a cluster's health, such as node status, number of unavailable ranges, and queries per second and service latency across the cluster. This tool is designed to help you optimize cluster performance and troubleshoot issues. The DB Console is accessible from every node at http://<host>:<http-port>, or http://<host>:8080 by default.

The DB Console automatically runs in the cluster. The following sections describe some of the pages that can help you to monitor and observe your cluster. For more information on accessing the DB Console, see Access DB Console.

Metrics dashboards

The Metrics dashboards, which are located within Metrics in DB Console, provide information about a cluster's performance, load, and resource utilization. The Metrics dashboards are built using time-series metrics collected from the cluster. By default, metrics are collected every 10 minutes and stored within the cluster, and data is retained at 10-second granularity for 10 days , and at 30-minute granularity for 90 days.

To learn more, refer to Overview Dashboard.

Each cluster automatically exposes its metrics at an endpoint in Prometheus format, enabling you to collect them in an external tool like Datadog or your own Prometheus, Grafana, and AlertManager instances. These tools:

• Collect metrics from the cluster's Prometheus endpoint at an interval you define.
• Store historical metrics according to your data retention requirements.
• Allow you to create and share dashboards, reports, and alerts based on metrics.
• Do not run within the cluster, and can help you to investigate a situation that led up to cluster outage even if the cluster is unavailable.

Metrics collected by the DB Console are stored within the cluster, and the SQL queries that create the reports on the Metrics dashboards also impose load on the cluster. To avoid this additional load, or if you rely on external tools for storing and visualizing your cluster's time-series metrics, Cockroach Labs recommends that you disable the DB Console's storage of time-series metrics.

When storage of time-series metrics is disabled, the cluster continues to expose its metrics via the Prometheus endpoint. The DB Console stops storing new time-series cluster metrics and eventually deletes historical data. The Metrics dashboards in the DB Console are still available, but their visualizations are blank. This is because the dashboards rely on data that is no longer available.

SQL Activity pages

The SQL Activity pages, which are located within SQL Activity in DB Console, provide information about SQL statements, transactions, and sessions.

The information on the SQL Activity pages comes from the cluster's crdb_internal system catalog. It is not exported via the cluster's Prometheus endpoint.

Cluster API

The Cluster API is a REST API that runs in the cluster and provides much of the same information about your cluster and nodes as is available from the DB Console or the Prometheus endpoint, and is accessible from each node at the same address and port as the DB Console.

If the cluster is unavailable, the Cluster API is also unavailable.

For more information, see the Cluster API overview and reference.

crdb_internal system catalog

The crdb_internal system catalog is a schema in each database that contains information about internal objects, processes, and metrics about that database. DBMarlin provides a third-party tool that collects metrics from a cluster's crdb_internal system catalogs rather than the cluster's Prometheus endpoint. Refer to Monitor CockroachDB with DBmarlin.

If the cluster is unavailable, a database's crdb_internal system catalog cannot be queried.

For details, see crdb_internal.

Health endpoints

CockroachDB provides two HTTP endpoints for checking the health of individual nodes.

These endpoints are also available through the Cluster API under /v2/health/.

If the cluster is unavailable, these endpoints are also unavailable.

/health

If a node is down, the http://<host>:<http-port>/health endpoint returns a Connection refused error:

$ curl http://localhost:8080/health

curl: (7) Failed to connect to localhost port 8080: Connection refused

Otherwise, it returns an HTTP 200 OK status response code with an empty body:

{

}

The /health endpoint does not returns details about the node such as its private IP address. These details could be considered privileged information in some deployments. If you need to retrieve node details, you can use the /_status/details endpoint along with a valid authentication cookie.

/health?ready=1

The http://<node-host>:<http-port>/health?ready=1 endpoint returns an HTTP 503 Service Unavailable status response code with an error in the following scenarios:

• The node is in the wait phase of the node shutdown sequence. This causes load balancers and connection managers to reroute traffic to other nodes before the node is drained of SQL client connections and leases, and is a necessary check during rolling upgrades.

  Tip:

  If you find that your load balancer's health check is not always recognizing a node as unready before the node shuts down, you can increase the server.shutdown.drain_wait cluster setting to cause a node to return 503 Service Unavailable even before it has started shutting down.

• The node is unable to communicate with a majority of the other nodes in the cluster, likely because the cluster is unavailable due to too many nodes being down.

$ curl http://localhost:8080/health?ready=1

{
  "error": "node is not healthy",
  "code": 14,
  "message": "node is not healthy",
  "details": [
  ]
}

Otherwise, it returns an HTTP 200 OK status response code with an empty body:

{

}

Raw status endpoints

Several endpoints return raw status metrics in JSON at http://<host>:<http-port>/#/debug. Feel free to investigate and use these endpoints, but note that they are subject to change.

Node status command

The cockroach node status command gives you metrics about the health and status of each node.

• With the --ranges flag, you get granular range and replica details, including unavailability and under-replication.
• With the --stats flag, you get granular disk usage details.
• With the --decommission flag, you get details about the node decommissioning process.
• With the --all flag, you get all of the above.

Prometheus endpoint

Every node of a CockroachDB cluster exports granular time-series metrics at http://<host>:<http-port>/_status/vars. The metrics are formatted for easy integration with Prometheus, an open source tool for storing, aggregating, and querying time-series data. The Prometheus format is human-readable and can be processed to work with other third-party monitoring systems such as Sysdig and stackdriver. Many of the third-party monitoring integrations, such as Datadog and Kibana, collect metrics from a cluster's Prometheus endpoint.

To access the Prometheus of a cluster running on localhost:8080:

$ curl http://localhost:8080/_status/vars

# HELP gossip_infos_received Number of received gossip Info objects
# TYPE gossip_infos_received counter
gossip_infos_received 0
# HELP sys_cgocalls Total number of cgo calls
# TYPE sys_cgocalls gauge
sys_cgocalls 3501
# HELP sys_cpu_sys_percent Current system cpu percentage
# TYPE sys_cpu_sys_percent gauge
sys_cpu_sys_percent 1.098855319644276e-10
# HELP replicas_quiescent Number of quiesced replicas
# TYPE replicas_quiescent gauge
replicas_quiescent{store="1"} 20
...

If you rely on external tools for storing and visualizing your cluster's time-series metrics, Cockroach Labs recommends that you disable the DB Console's storage of time-series metrics.

When storage of time-series metrics is disabled, the DB Console Metrics dashboards in the DB Console are still available, but their visualizations are blank. This is because the dashboards rely on data that is no longer available.

Critical nodes endpoint

The critical nodes status endpoint is used to:

• Check if any of your nodes are in a critical state. A node is critical if that node becoming unreachable would cause replicas to become unavailable.
• Check if any ranges are under-replicated or unavailable. This is useful when determining whether a node is ready for decommissioning.
• Check if any of your cluster's data placement constraints (set via multi-region SQL or direct configuration of replication zones) are being violated. This is useful when implementing data domiciling.

Fields

The JSON object returned by the critical nodes status endpoint contains the following top-level fields.

The criticalNodes portion of the response contains a (possibly empty) list of objects, each of which has the following fields.

Each report subtype (e.g., report.unavailable, report.violatingConstraints, etc.) returns a (possibly empty) list of objects describing the ranges that report applies to. Each object contains a rangeDescriptor and a config that describes the range.

Examples

• Replication status - normal
• Replication status - constraint violation
• Replication status - under-replicated ranges
• Replication status - ranges in critical localities

Replication status - normal

The following example assumes you are running a newly started, local cockroach demo cluster started using the following command:

cockroach demo --geo-partitioned-replicas --insecure

curl -X POST http://localhost:8080/_status/critical_nodes

{
  "criticalNodes": [
  ],
  "report": {
    "underReplicated": [
    ],
    "overReplicated": [
    ],
    "violatingConstraints": [
    ],
    "unavailable": [
    ],
    "unavailableNodeIds": [
    ]
  }
}

Replication status - constraint violation

The following example assumes you are running a newly started, local cockroach demo cluster started using the following command:

cockroach demo --geo-partitioned-replicas --insecure

By default, this geo-distributed demo cluster will not have any constraint violations.

To introduce a violation that you can then query for, you'll modify the zone configuration of the movr.users table.

You can use SHOW CREATE TABLE to see what existing zone configurations are attached to the users table, so you know what to modify.

SHOW CREATE TABLE users;

  table_name |                                    create_statement
-------------+-----------------------------------------------------------------------------------------
  users      | CREATE TABLE public.users (
             |     id UUID NOT NULL,
             |     city VARCHAR NOT NULL,
             |     name VARCHAR NULL,
             |     address VARCHAR NULL,
             |     credit_card VARCHAR NULL,
             |     CONSTRAINT users_pkey PRIMARY KEY (city ASC, id ASC)
             | ) PARTITION BY LIST (city) (
             |     PARTITION us_west VALUES IN (('seattle'), ('san francisco'), ('los angeles')),
             |     PARTITION us_east VALUES IN (('new york'), ('boston'), ('washington dc')),
             |     PARTITION europe_west VALUES IN (('amsterdam'), ('paris'), ('rome'))
             | );
             | ALTER PARTITION europe_west OF INDEX movr.public.users@users_pkey CONFIGURE ZONE USING
             |     constraints = '[+region=europe-west1]';
             | ALTER PARTITION us_east OF INDEX movr.public.users@users_pkey CONFIGURE ZONE USING
             |     constraints = '[+region=us-east1]';
             | ALTER PARTITION us_west OF INDEX movr.public.users@users_pkey CONFIGURE ZONE USING
             |     constraints = '[+region=us-west1]'
(1 row)

To create a constraint violation, use the ALTER PARTITION statement to tell the ranges in the europe_west partition that they are explicitly supposed to not be in the region=europe-west1 locality:

ALTER PARTITION europe_west of INDEX movr.public.users@users_pkey CONFIGURE ZONE USING constraints = '[-region=europe-west1]';

Once the statement above executes, the ranges currently stored in that locality will now be in a state where they are explicitly not supposed to be in that locality, and are thus in violation of a constraint.

In other words, this tells the ranges that "where you are now is exactly where you are not supposed to be". This will cause the cluster to rebalance the ranges, which will take some time. During the time it takes for the rebalancing to occur, the ranges will be in violation of a constraint.

The critical nodes endpoint should now report a constraint violation in the violatingConstraints field of the response, similar to the one shown below.

curl -X POST http://localhost:8080/_status/critical_nodes

{
  "criticalNodes": [
  ],
  "report": {
    "underReplicated": [
    ],
    "overReplicated": [
    ],
    "violatingConstraints": [
      {
        "rangeDescriptor": {
          "rangeId": "89",
          "startKey": "8okSYW1zdGVyZGFtAAE=",
          "endKey": "8okSYW1zdGVyZGFtAAESszMzMzMzQAD/gAD/AP8A/wD/AP8A/yMAAQ==",
          "internalReplicas": [
            {
              "nodeId": 8,
              "storeId": 8,
              "replicaId": 5,
              "type": 0
            },
            {
              "nodeId": 4,
              "storeId": 4,
              "replicaId": 7,
              "type": 0
            },
            {
              "nodeId": 3,
              "storeId": 3,
              "replicaId": 6,
              "type": 0
            }
          ],
          "nextReplicaId": 8,
          "generation": "40",
          "stickyBit": {
            "wallTime": "0",
            "logical": 0,
            "synthetic": false
          }
        },
        "config": {
          "rangeMinBytes": "134217728",
          "rangeMaxBytes": "536870912",
          "gcPolicy": {
            "ttlSeconds": 14400,
            "protectionPolicies": [
            ],
            "ignoreStrictEnforcement": false
          },
          "globalReads": false,
          "numReplicas": 3,
          "numVoters": 0,
          "constraints": [
            {
              "numReplicas": 0,
              "constraints": [
                {
                  "type": 1,
                  "key": "region",
                  "value": "europe-west1"
                }
              ]
            }
          ],
          "voterConstraints": [
          ],
          "leasePreferences": [
          ],
          "rangefeedEnabled": false,
          "excludeDataFromBackup": false
        }
      },
      ...
    ],
    "unavailable": [
    ],
    "unavailableNodeIds": [
    ]
  }
}

Replication status - under-replicated ranges

The following example assumes you are running a newly started, local cockroach demo cluster started using the following command:

cockroach demo --geo-partitioned-replicas --insecure

By default, this geo-distributed demo cluster will not have any under-replicated ranges.

To put the cluster into a state where some ranges are under-replicated, issue the following ALTER TABLE ... CONFIGURE ZONE statement, which tells it to store 9 copies of each range underlying the rides table.

ALTER TABLE rides CONFIGURE ZONE USING num_replicas=9;

Once the statement above executes, the cluster will rebalance so that it's storing 9 copies of each range underlying the rides table. During the time it takes for the rebalancing to occur, these ranges will be considered under-replicated, since there are not yet as many copies (9) of each range as you have just specified.

The critical nodes endpoint should now report a constraint violation in the underReplicated field of the response.

curl -X POST http://localhost:8080/_status/critical_nodes

{
  "criticalNodes": [
    {
      "nodeId": 2,
      "address": {
        "networkField": "tcp",
        "addressField": "127.0.0.1:26358"
      },
      "attrs": {
        "attrs": [
        ]
      },
      "locality": {
        "tiers": [
          {
            "key": "region",
            "value": "us-east1"
          },
          {
            "key": "az",
            "value": "c"
          }
        ]
      },
      "ServerVersion": {
        "majorVal": 23,
        "minorVal": 1,
        "patch": 0,
        "internal": 0
      },
      "buildTag": "v23.1.0-rc.2",
      "startedAt": "1683655799845426000",
      "localityAddress": [
      ],
      "clusterName": "",
      "sqlAddress": {
        "networkField": "tcp",
        "addressField": "127.0.0.1:26258"
      },
      "httpAddress": {
        "networkField": "tcp",
        "addressField": "127.0.0.1:8081"
      }
    },
    ...
  ],
  "report": {
    "underReplicated": [
      {
        "rangeDescriptor": {
          "rangeId": "76",
          "startKey": "9A==",
          "endKey": "9IkSYW1zdGVyZGFtAAE=",
          "internalReplicas": [
            {
              "nodeId": 3,
              "storeId": 3,
              "replicaId": 4,
              "type": 0
            },
            {
              "nodeId": 4,
              "storeId": 4,
              "replicaId": 2,
              "type": 0
            },
            {
              "nodeId": 8,
              "storeId": 8,
              "replicaId": 3,
              "type": 0
            }
          ],
          "nextReplicaId": 5,
          "generation": "44",
          "stickyBit": {
            "wallTime": "0",
            "logical": 0,
            "synthetic": false
          }
        },
        "config": {
          "rangeMinBytes": "134217728",
          "rangeMaxBytes": "536870912",
          "gcPolicy": {
            "ttlSeconds": 14400,
            "protectionPolicies": [
            ],
            "ignoreStrictEnforcement": false
          },
          "globalReads": false,
          "numReplicas": 9,
          "numVoters": 0,
          "constraints": [
          ],
          "voterConstraints": [
          ],
          "leasePreferences": [
          ],
          "rangefeedEnabled": false,
          "excludeDataFromBackup": false
        }
      },
      ...
    ],
    "overReplicated": [
    ],
    "violatingConstraints": [
      {
        "rangeDescriptor": {
          "rangeId": "238",
          "startKey": "9IkSYW1zdGVyZGFtAAE=",
          "endKey": "9IkSYW1zdGVyZGFtAAESxR64UeuFQAD/gAD/AP8A/wD/AP8BgQAB",
          "internalReplicas": [
            {
              "nodeId": 9,
              "storeId": 9,
              "replicaId": 5,
              "type": 0
            },
            {
              "nodeId": 7,
              "storeId": 7,
              "replicaId": 4,
              "type": 0
            },
            {
              "nodeId": 8,
              "storeId": 8,
              "replicaId": 3,
              "type": 0
            }
          ],
          "nextReplicaId": 6,
          "generation": "48",
          "stickyBit": {
            "wallTime": "0",
            "logical": 0,
            "synthetic": false
          }
        },
        "config": {
          "rangeMinBytes": "134217728",
          "rangeMaxBytes": "536870912",
          "gcPolicy": {
            "ttlSeconds": 14400,
            "protectionPolicies": [
            ],
            "ignoreStrictEnforcement": false
          },
          "globalReads": false,
          "numReplicas": 9,
          "numVoters": 0,
          "constraints": [
            {
              "numReplicas": 0,
              "constraints": [
                {
                  "type": 0,
                  "key": "region",
                  "value": "europe-west1"
                }
              ]
            }
          ],
          "voterConstraints": [
          ],
          "leasePreferences": [
          ],
          "rangefeedEnabled": false,
          "excludeDataFromBackup": false
        }
      },
      ...
    ],
    "unavailable": [
    ],
    "unavailableNodeIds": [
    ]
  }
}

Replication status - ranges in critical localities

The following example assumes you are running a newly started, local cockroach demo cluster started using the following command:

cockroach demo --geo-partitioned-replicas --insecure

By default, this geo-distributed demo cluster will not have any nodes in a critical state. A node is critical if that node becoming unreachable would cause replicas to become unavailable.

The status endpoint describes which of your nodes (if any) are critical via the criticalNodes field in the response.

To artificially put the nodes in this demo cluster in "critical" status, we can issue the following SQL statement, which uses ALTER TABLE ... CONFIGURE ZONE to tell the cluster to store more copies of each range underlying the rides table than there are nodes in the cluster.

ALTER TABLE rides CONFIGURE ZONE USING num_replicas=128;

The critical nodes endpoint should now report that all of the cluster's nodes are critical by listing them in the criticalNodes field of the response.

curl -X POST http://localhost:8080/_status/critical_nodes

{
  "criticalNodes": [
    {
      "nodeId": 4,
      "address": {
        "networkField": "tcp",
        "addressField": "127.0.0.1:26360"
      },
      "attrs": {
        "attrs": [
        ]
      },
      "locality": {
        "tiers": [
          {
            "key": "region",
            "value": "us-west1"
          },
          {
            "key": "az",
            "value": "a"
          }
        ]
      },
      "ServerVersion": {
        "majorVal": 23,
        "minorVal": 1,
        "patch": 0,
        "internal": 0
      },
      "buildTag": "v23.1.0-rc.2",
      "startedAt": "1683656700210217000",
      "localityAddress": [
      ],
      "clusterName": "",
      "sqlAddress": {
        "networkField": "tcp",
        "addressField": "127.0.0.1:26260"
      },
      "httpAddress": {
        "networkField": "tcp",
        "addressField": "127.0.0.1:8083"
      }
    },
    ...
  ],
  "report": {
    "underReplicated": [
      {
        "rangeDescriptor": {
          "rangeId": "133",
          "startKey": "9A==",
          "endKey": "9IkSYW1zdGVyZGFtAAE=",
          "internalReplicas": [
            {
              "nodeId": 1,
              "storeId": 1,
              "replicaId": 5,
              "type": 0
            },
            {
              "nodeId": 4,
              "storeId": 4,
              "replicaId": 6,
              "type": 0
            },
            {
              "nodeId": 8,
              "storeId": 8,
              "replicaId": 3,
              "type": 0
            }
          ],
          "nextReplicaId": 7,
          "generation": "52",
          "stickyBit": {
            "wallTime": "0",
            "logical": 0,
            "synthetic": false
          }
        },
        "config": {
          "rangeMinBytes": "134217728",
          "rangeMaxBytes": "536870912",
          "gcPolicy": {
            "ttlSeconds": 14400,
            "protectionPolicies": [
            ],
            "ignoreStrictEnforcement": false
          },
          "globalReads": false,
          "numReplicas": 128,
          "numVoters": 0,
          "constraints": [
          ],
          "voterConstraints": [
          ],
          "leasePreferences": [
          ],
          "rangefeedEnabled": false,
          "excludeDataFromBackup": false
        }
      },
      ...
    ],
    "overReplicated": [
    ],
    "violatingConstraints": [
      {
        "rangeDescriptor": {
          "rangeId": "186",
          "startKey": "9IkSYW1zdGVyZGFtAAE=",
          "endKey": "9IkSYW1zdGVyZGFtAAESxR64UeuFQAD/gAD/AP8A/wD/AP8BgQAB",
          "internalReplicas": [
            {
              "nodeId": 7,
              "storeId": 7,
              "replicaId": 5,
              "type": 0
            },
            {
              "nodeId": 9,
              "storeId": 9,
              "replicaId": 4,
              "type": 0
            },
            {
              "nodeId": 8,
              "storeId": 8,
              "replicaId": 3,
              "type": 0
            }
          ],
          "nextReplicaId": 6,
          "generation": "52",
          "stickyBit": {
            "wallTime": "0",
            "logical": 0,
            "synthetic": false
          }
        },
        "config": {
          "rangeMinBytes": "134217728",
          "rangeMaxBytes": "536870912",
          "gcPolicy": {
            "ttlSeconds": 14400,
            "protectionPolicies": [
            ],
            "ignoreStrictEnforcement": false
          },
          "globalReads": false,
          "numReplicas": 128,
          "numVoters": 0,
          "constraints": [
            {
              "numReplicas": 0,
              "constraints": [
                {
                  "type": 0,
                  "key": "region",
                  "value": "europe-west1"
                }
              ]
            }
          ],
          "voterConstraints": [
          ],
          "leasePreferences": [
          ],
          "rangefeedEnabled": false,
          "excludeDataFromBackup": false
        }
      },
      ...
    ],
    "unavailable": [
    ],
    "unavailableNodeIds": [
    ]
  }
}

Alerting tools

In addition to actively monitoring the overall health and performance of a cluster, it is also essential to configure alerting rules that promptly send notifications when CockroachDB experiences events that require investigation or intervention.

Many of the third-party monitoring integrations, such as Datadog and Kibana, also support event-based alerting using metrics collected from a cluster's Prometheus endpoint. Refer to the documentation for an integration for more details. This section identifies the most important events that you might want to create alerting rules for, and provides pre-defined rules definitions for these events appropriate for use with Prometheus's open source Alertmanager service.

Alertmanager

If you have configured Prometheus to monitor your CockroachDB instance, you can also configure alerting rule definitions to have Alertmanager detect important events and alert you when they occur.

If you rely on external tools for storing and visualizing your cluster's time-series metrics, Cockroach Labs recommends that you disable the DB Console's storage of time-series metrics.

When storage of time-series metrics is disabled, the DB Console Metrics dashboards in the DB Console are still available, but their visualizations are blank. This is because the dashboards rely on data that is no longer available.

Prometheus alerting rules endpoint

Every CockroachDB node exports an alerting rules template at http://<host>:<http-port>/api/v2/rules/. These rule definitions are formatted for easy integration with Alertmanager.

$ curl http://localhost:8080/api/v2/rules/

rules/alerts:
    rules:
        - alert: UnavailableRanges
          expr: (sum by(instance, cluster) (ranges_unavailable)) > 0
          for: 10m0s
          annotations:
            summary: Instance {{ $labels.instance }} has {{ $value }} unavailable ranges
        - alert: TrippedReplicaCircuitBreakers
          expr: (sum by(instance, cluster) (kv_replica_circuit_breaker_num_tripped_replicas)) > 0
          for: 10m0s
          annotations:
            summary: Instance {{ $labels.instance }} has {{ $value }} tripped per-Replica circuit breakers
...

Working with Alertmanager rules

To add a rule from the api/v2/rules/ rules endpoint, create or edit your alerts.rules.yml file and copy the rule definition for the event you want to alert on. For example, to add a rule to alert you when unavailable ranges are detected, copy the following from the rules endpoint into your alerts.rules.yml file:

- alert: UnavailableRanges
  expr: (sum by(instance, cluster) (ranges_unavailable)) > 0
  for: 10m0s
  annotations:
    summary: Instance {{ $labels.instance }} has {{ $value }} unavailable ranges

If you already followed the steps from Monitor CockroachDB with Prometheus, you should already have a alerts.rules.yml file. If you are creating a new alerts.rules.yml file, be sure that it begins with the following three lines:

groups:
- name: rules/alerts.rules
  rules:

Place your desired rule(s) underneath the rules: header. For example, the following shows an alerts.rules.yml file with the unavailable ranges rule defined:

groups:
- name: rules/alerts.rules
  rules:
  - alert: UnavailableRanges
    expr: (sum by(instance, cluster) (ranges_unavailable)) > 0
    for: 10m0s
    annotations:
      summary: Instance {{ $labels.instance }} has {{ $value }} unavailable ranges

Once you have created or edited your alerts.rules.yml file, reference it in your prometheus.yml configuration file with the following:

rule_files:
- "rules/alerts.rules.yml"

If you already followed the steps from Monitor CockroachDB with Prometheus, this reference is already present in your prometheus.yml file.

Start Prometheus and Alertmanager to begin watching for events to alert on. You can view imported rules on your Prometheus server's web interface at http://<host>:<http-port>/rules. Use the "State" column to verify that the rules were imported correctly.

Events to alert on

Node is down

• Rule: Send an alert when a node has been down for 15 minutes or more.

• How to detect: If a node is down, its _status/vars endpoint will return a Connection refused error. Otherwise, the liveness_livenodes metric will be the total number of live nodes in the cluster.

• Rule definition: Use the InstanceDead alert from our pre-defined alerting rules.

Node is restarting too frequently

• Rule: Send an alert if a node has restarted more than once in the last 10 minutes.

• How to detect: Calculate this using the number of times the sys_uptime metric in the node's _status/vars output was reset back to zero. The sys_uptime metric gives you the length of time, in seconds, that the cockroach process has been running.

• Rule definition: Use the InstanceFlapping alert from our pre-defined alerting rules.

Node is running low on disk space

• Rule: Send an alert when a node has less than 15% of free space remaining.

• How to detect: Divide the capacity metric by the capacity_available metric in the node's _status/vars output.

• Rule definition: Use the StoreDiskLow alert from our pre-defined alerting rules.

For instructions on how to free up disk space as quickly as possible after deleting data, see How can I free up disk space quickly?

Node is not executing SQL

• Rule: Send an alert when a node is not executing SQL despite having connections.

• How to detect: The sql_conns metric in the node's _status/vars output will be greater than 0 while the sql_query_count metric will be 0. You can also break this down by statement type using sql_select_count, sql_insert_count, sql_update_count, and sql_delete_count.

CA certificate expires soon

• Rule: Send an alert when the CA certificate on a node will expire in less than a year.

• How to detect: Calculate this using the security_certificate_expiration_ca metric in the node's _status/vars output.

• Rule definition: Use the CACertificateExpiresSoon alert from our pre-defined alerting rules.

Node certificate expires soon

• Rule: Send an alert when a node's certificate will expire in less than a year.

• How to detect: Calculate this using the security_certificate_expiration_node metric in the node's _status/vars output.

• Rule definition: Use the NodeCertificateExpiresSoon alert from our pre-defined alerting rules.

Changefeed is experiencing high latency

• Rule: Send an alert when the latency of any changefeed running on any node is higher than the set threshold, which depends on the gc.ttlseconds variable set in the cluster.

• How to detect: Calculate this using a threshold, where the threshold is less than the value of the gc.ttlseconds variable. For example, changefeed.max_behind_nanos > [some threshold].

Unavailable ranges

• Rule: Send an alert when the number of ranges with fewer live replicas than needed for quorum is non-zero for too long.

• How to detect: Calculate this using the ranges_unavailable metric in the node's _status/vars output.

• Rule definition: Use the UnavailableRanges alerting rule from your cluster's api/v2/rules/ metrics endpoint.

Tripped replica circuit breakers

• Rule: Send an alert when a replica stops serving traffic due to other replicas being offline for too long.

• How to detect: Calculate this using the kv_replica_circuit_breaker_num_tripped_replicas metric in the node's _status/vars output.

• Rule definition: Use the TrippedReplicaCircuitBreakers alerting rule from your cluster's api/v2/rules/ metrics endpoint.

Under-replicated ranges

• Rule: Send an alert when the number of ranges with replication below the replication factor is non-zero for too long.

• How to detect: Calculate this using the ranges_underreplicated metric in the node's _status/vars output.

• Rule definition: Use the UnderreplicatedRanges alerting rule from your cluster's api/v2/rules/ metrics endpoint.

Requests stuck in Raft

• Rule: Send an alert when requests are taking a very long time in replication.

• How to detect: Calculate this using the requests_slow_raft metric in the node's _status/vars output.

• Rule definition: Use the RequestsStuckInRaft alerting rule from your cluster's api/v2/rules/ metrics endpoint.

High open file descriptor count

• Rule: Send an alert when a cluster is getting close to the open file descriptor limit.

• How to detect: Calculate this using the sys_fd_softlimit metric in the node's _status/vars output.

• Rule definition: Use the HighOpenFDCount alerting rule from your cluster's api/v2/rules/ metrics endpoint.

See also

• Production Checklist
• Manual Deployment
• Orchestrated Deployment
• Local Deployment
• Third-Party Monitoring Integrations
• Metrics

Was this helpful?