Connecting to 192.168.1.100:3128... failed: No route to host.

[Dan Prince]

Today CI jobs are failing due to the following:

2015-06-15 16:51:08.937 | + '[' -n '' ']'
2015-06-15 16:51:08.937 | + FEDORA_IMAGE=fedora-21.x86_64.qcow2
2015-06-15 16:51:08.938 | + wget --progress=dot:mega http://cloud.fedoraproject.org/fedora-21.x86_64.qcow2
2015-06-15 16:51:08.940 | --2015-06-15 16:51:08-- http://cloud.fedoraproject.org/fedora-21.x86_64.qcow2
2015-06-15 16:51:11.931 | Connecting to 192.168.1.100:3128... failed: No route to host.

[Dan Prince (dan-prince)]

Looking into things... 192.168.1.100 is our local squid proxy. Turns out the compute host for this machine had a kernel panic and was offline.

A couple more compute hosts seem to be suffering similar issues so I'm looking into what could be the issue.

[Dan Prince (dan-prince)]

The squid is back online again after a full rebuild of the compute hosts.

The root cause of the CI outage turned out to be that neutron-server was unable to authenticate users against keystone due to a bad keystone-signing directory.