memory leak in grafana 4.1.2 + hawkular-datasource 1.0.7 + hawkular 0.18

[fche]

A pretty vanilla dockerfile based on the docker/* sample here, deployed to openshift 3.3.1.3, results in a steady memory leak. The only activity to the grafana server is a periodic (10s) liveness test to the / URL, so am not sure that the hawkular datasource is even implicated. I'm reporting here first to see if this sounds familiar.

/proc/.../maps after 12 hours of runtime
http://sprunge.us/PQOE

OS console metrics over same 12 hours

[jotak]

If I understand, your "osd-monitor" container is really just Grafana + this datasource plugin?

[fche]

yup
wondering how to even start diagnosing such problems

[fche]

Later testing seems to implicate the periodic health-testing inquiries from openshift. Is there any plausible hawkular-datasource linkage to this?

      livenessProbe:
        failureThreshold: 3
        httpGet:
          path: /
          port: 3000
          scheme: HTTP
        initialDelaySeconds: 1
        periodSeconds: 10
        successThreshold: 1
        timeoutSeconds: 1
      readinessProbe:
        [same]