Uploaded image for project: 'Cloud Enablement'
  1. Cloud Enablement
  2. CLOUD-2996

[EAP][CD][7.2] scale up, new EAP pod ready before it joined the cluster, sessions lost during merge


    • CLOUD Maintenance Sprint 27

      With current EAP builds (EAP CD, EAP 7.2) When scaling up (under load), the time it takes the new node to join the cluster seems to be significantly higher that it is with jboss-eap-7/eap71-openshift:1.3.

      More imporantly, in our test scenario, the new pod is ready before it joins the cluster (so it is receiving traffic) and the sessions such initiated are lost when the node joins the rest of the cluster.

      Scenario: simple <distributable/> servlet, storing an integer counter into session
      scaling from 2 to 3 under load (256 clients with random 0-20s delay before each request), no encryption, openshift.DNS_PING

        1. 3pods-dns.zip
          63 kB
        2. CLOUD-2996.tar.gz
          1.91 MB
        3. CLOUD-3016.tar.bz2
          20.28 MB
        4. CLOUD-3016-GMS30TRACE.dns.DNS.dockerKillOneOfThree-1.tar.bz2
          18.77 MB
        5. dockerKillOneOfThree.png
          93 kB
        6. dockerKillOneOfThree.tar.gz
          1.87 MB
        7. jgroups-good.yaml
          0.9 kB
        8. scaleUpFromTwoToThree.png
          101 kB
        9. standalone-openshift.xml
          36 kB

            kwills@redhat.com Ken Wills
            maschmid@redhat.com Marek Schmidt
            Marek Schmidt Marek Schmidt
            0 Vote for this issue
            11 Start watching this issue