These are the key performance indicators for the Fili Web Service component, listed in categories by order of importance.
Server Error Responses (HTTP 5XX)
Shows how much trouble the service is having.
- com.codahale.metrics.servlet.AbstractInstrumentedFilter.responseCodes.serverError.m1_rate
Druid Errors
Shows how much trouble queries are having against druid.
- druid.errors.exceptions.m1_rate
- druid.errors.http.m1_rate
Requests
Shows how many requests the service is serving.
- com.codahale.metrics.servlet.AbstractInstrumentedFilter.requests.m1_rate
- com.codahale.metrics.servlet.AbstractInstrumentedFilter.requests.m15_rate
System Metrics
Shows the overall health of the system’s low-level resources and activities.
- CPU
- Memory
- Network IO
- GC Pauses
Latency
Shows duration of overall requests and druid requests. (m1_rate and pN)
- com.codahale.metrics.servlet.AbstractInstrumentedFilter.requests.p50
- com.codahale.metrics.servlet.AbstractInstrumentedFilter.requests.p75
- com.codahale.metrics.servlet.AbstractInstrumentedFilter.requests.p95
- com.codahale.metrics.servlet.AbstractInstrumentedFilter.requests.p98
- com.codahale.metrics.servlet.AbstractInstrumentedFilter.requests.p99
- com.codahale.metrics.servlet.AbstractInstrumentedFilter.requests.p999
- druid.requests.p50
- druid.requests.p75
- druid.requests.p95
- druid.requests.p98
- druid.requests.p99
- druid.requests.p999
Rate Limiting Rejections
Shows if users are hitting rate limits.
- ratelimit.meter.reject.ui.m1_rate
- ratelimit.meter.reject.user.m1_rate
- ratelimit.meter.reject.global.m1_rate
Active Requests
Shows load at a given point in time. (ie. how close are the load is to the limits of Druid)
- com.codahale.metrics.servlet.AbstractInstrumentedFilter.activeRequests.count
Bad Request Responses (HTTP 4XX)
Shows how much trouble users are having interacting with the API.
- com.codahale.metrics.servlet.AbstractInstrumentedFilter.responseCodes.badRequest.m1_rate
- com.codahale.metrics.servlet.AbstractInstrumentedFilter.responseCodes.notFound.m1_rate