Comment 0 for bug 1255242

Revision history for this message
Curtis Hovey (sinzui) wrote : upgrade-juju on HP cloud broken in devel

I am sorry for this late bug report. CI consistently shows that upgrade-juju consistently fails on hp-cloud. Revno 2072 appears to be the culprit. Revno 2071 was the last pass, and we verified 2071 was good in two repeats runs of CI. Revno 2072 always fails and fails the same way...the mysql charm has a hook error.

The specific test that fails does this: A 1.16.3 env is bootstrapped on hp cloud. Wordpress + mysql is deployed and verified to be up. The test workspace installed the juju-core 1.17.0. package created from revno 2072. The new tools and metadata were published to all the clouds uder the testing/ prefix. Next juju-upgrade was called on HP cloud and we know 1.17.0 was found.

We see this error after the second effort to get the tools
    ERROR juju runner.go:220 worker: exited "environ-provisioner": no state server machines with addresses found

Much latter we see a hook failure
    juju-test-release-hp-machine-2:2013-11-26 15:10:05 INFO juju.worker.uniter context.go:263
    HOOK subprocess.CalledProcessError: Command '['service', 'mysql', 'start']' returned non-zero exit status 1

The mysql charm did not fail on aws, canonistack, lxc, or azure. Just Hp.

I am attaching the juju and CI log.