ContextQMD
Back to Nightingale

TiDB

integrations/TiDB/markdown/README.en_US.md

9.1.11.3 KB
Original Source

TiDB

TiDB, PD, and TiKV expose Prometheus metrics through HTTP:

See the TiDB Monitoring API.

ComponentDefault endpoint
TiDB Serverhttp://<tidb>:10080/metrics
PD Serverhttp://<pd>:2379/metrics
TiKV Serverhttp://<tikv>:20180/metrics

The dashboard depends on k8s_cluster, tidb_cluster, job, and instance. Set stable cluster labels even for a non-Kubernetes deployment:

toml
[[instances]]
urls = ["http://tidb-01:10080/metrics"]
url_label_key = "instance"
url_label_value = "{{.Host}}"
labels = { job = "tidb", tidb_cluster = "prod-tidb", k8s_cluster = "baremetal" }

[[instances]]
urls = ["http://pd-01:2379/metrics"]
url_label_key = "instance"
url_label_value = "{{.Host}}"
labels = { job = "pd", tidb_cluster = "prod-tidb", k8s_cluster = "baremetal" }

[[instances]]
urls = ["http://tikv-01:20180/metrics"]
url_label_key = "instance"
url_label_value = "{{.Host}}"
labels = { job = "tikv", tidb_cluster = "prod-tidb", k8s_cluster = "baremetal" }

Add every TiDB, PD, and TiKV node. Scrape TiFlash, TiCDC, and other optional components only when deployed. The overview's probe_success panels also require black-box probes with matching cluster and group labels.