good first issuehelm
Repository metrics
- Stars
- (5,181 stars)
- PR merge metrics
- (PR metrics pending)
Description
The bundled nginx that comes in the helm chart doesn't expose all debug endpoints, only some. As such operators have to port-forward to a specific pod and open their browser. This can create confusion because the operator needs to know which pod to port-forward to and can be inconvenient.
For reference these are the currently exposed endpoints
| endpoint | component | done |
|---|---|---|
GET /distributor/all_user_stats |
distributor | |
GET /distributor/ring |
distributor | |
GET /distributor/ha_tracker |
distributor | |
GET /multitenant_alertmanager/status |
alertmanager |
I propose to expose these endpoints in addition to the 4 above
| endpoint | component | done in |
|---|---|---|
GET /memberlist |
query-frontend (doesn't matter) | |
GET /config |
query-frontend (doesn't matter) | |
GET /runtime_config |
query-frontend (doesn't matter) | |
GET /api/v1/user_limits |
query-frontend | |
GET /ruler/rule_groups |
ruler | |
GET /multitenant_alertmanager/configs |
alertmanager | https://github.com/grafana/mimir/pull/8248 |
GET /multitenant_alertmanager/ring |
alertmanager | |
GET /store-gateway/ring |
store-gateway | |
GET /store-gateway/tenants |
store-gateway | |
GET /store-gateway/tenant/{tenant}/blocks |
store-gateway | |
GET /compactor/ring |
compactor | |
GET /overrides-exporter/ring |
overrides-exporter | |
GET /ingester/ring |
ingester | |
GET /ingester/tsdb_metrics |
ingester |
and we probably don't want to expose
/compactor/delete_tenant, /compactor/delete_tenant_status, /ingester/flush, /ingester/shutdown, and /ingester/prepare-shutdown
These endpoints should be exposed on the nginx, gateway nginx, and the root-level ingress