I noticed a weird behavior of Spinnaker 1.6.0 running on GKE. When Spinnaker is deployed via halyard 0.42.0, the new versions of pods start to replace the old ones (for example spin-orca-v001-ckctx to replace spin-orca-v000-vb7cs) as shown below:
spin-orca-v000-vb7cs 0/2 Terminating 0 1h spin-orca-v001-ckctx 2/2 Running 0 3m
But I noticed that after some time the old version of Orca is restarted for some reason:
spin-orca-v000-xdd2p 1/2 Running 0 20s spin-orca-v001-ckctx 2/2 Running 0 4m
As a result the deployment takes too much time and is timed out in the end.
I noticed the same behavior when I tried to deploy an application using Spinnaker 1.6.0 using a simple pipeline with “highlander” strategy. So this looks like an issue in Clouddriver/Orca rather than in Halyard.
I didn’t observe this issue for Spinnaker 1.5.4.
@lwander is this a bug or I have some misconfiguration?