What is the time frame used by the Apigee-generated graphite FLOW percentile metrics? It seems like it's hours rather than what I would expect: since the last $reporting.interval (60 seconds by default). Details below:
We're utilizing the built-in graphite metrics with OPDK 4.15.7.00 which are very helpful though I haven't been able to find good documentation on them.
The message-processor FLOW domain generates metrics like:
${graphite.prefix}.FLOW.proxies.$ORG.$ENV.$PROXY.$REVISION.proxyLatency.95percentile
${graphite.prefix}.FLOW.proxies.$ORG.$ENV.$PROXY.$REVISION.targetLatency.95percentile
These values seem to to be measured in nanoseconds.
We had a back-end target perform slowly for about 30 minutes and then it was restored to normal. But, these metrics stayed elevated for hours afterwards though all indications showed that the service was responding quickly (e.g., access logs, analytics database).
LinkedIn
Twitter
