Uploaded image for project: 'OpenShift Bugs'
  1. OpenShift Bugs
  2. OCPBUGS-41902

No graceful shutdown in our OCP router pod.

XMLWordPrintable

    • Icon: Bug Bug
    • Resolution: Not a Bug
    • Icon: Major Major
    • None
    • 4.13.z, 4.12.z, 4.14.z, 4.15.z, 4.16.z
    • Networking / router
    • None
    • Moderate
    • None
    • 1
    • NE Sprint 259
    • 1
    • Rejected
    • False
    • Hide

      None

      Show
      None

      There's no `preStop` hook in current router pod, which means haproxy process will receive a `SIGTERM` right after OCP tries terminate it and after a 45 second wait. , the router pod finish termination. 

      But, according to HAProxy's document. Sending `SIGTERM` to haproxy process is a hard stop.

      So, we should send `SIGUSR1` instead.

       

      Here's a preStop hook example to send `SIGUSR1` to haproxy process.

            preStop:
              exec:
                command: ["/bin/sh", "-c", "kill -SIGUSR1 $(cat /var/lib/haproxy/run/haproxy.pid)"]

       

      Also, please consider allow user to define global grace option in haproxy. Because it is a important option in haproxy soft shutdown. 

      https://github.com/kaspergrubbe/haproxy-grace-test

              mmasters1@redhat.com Miciah Masters
              rhn-support-jaliang Jace Liang
              Hongan Li Hongan Li
              Jace Liang
              Votes:
              2 Vote for this issue
              Watchers:
              5 Start watching this issue

                Created:
                Updated:
                Resolved: