POD status isn't updating properly

[hodrigohamalho]

Hexboard and sketchpod

The pod sketchpod-1-sihq9 is on live state:

Checkin this pod (sketchpod-1-sihq9 ), it seems to be responding

[ryanj]

What method did you use to set up OpenShift?

How did you set up the app?

I'm not currently getting this issue in my 3.1 environments, but I've heard a few other people report similar issues.

[hodrigohamalho]

I used "quick install"

atomic-openshift-installer install

My environment is on 2 virtual machines on RHEL 7.2.

To setup the application I used:

oc new-app -f https://raw.githubusercontent.com/2015-Middleware-Keynote/hexboard/master/app_template.json -p ACCESS_TOKEN=$(oc whoami -t)

after that scalled sketchpod via GUI.

[detiber]

I can confirm this... From what I could gather it looks like the initial list/watch is working successfully (anything that completes within that timeframe continues to work), but subsequent watches do not appear to be occuring.

As an aside, my environment was provisioned with demo-ansible (albeit with the ha changes I have been working on).

[thoraxe]

I can confirm this is still kinda happening. I scaled 1->10 and then 10->63 and was left with 2 pods that were running but did not show on the hexboard. I quickly scanned the hexboard log and found no errors. I just built the hexboard with the latest master code (14-JAN 05:58AM EST) which incorporates the merged changes from 13-JAN.

[ryanj]

Thanks for testing. A few more patches have been merged.

Let me know if you're still seeing the same issues.

[hodrigohamalho]

Still doesn't work here.

[ryanj]

I'm getting better results with the origin-1.1.3 vm from openshift.org/vm. Auto-recovery seems to be working, but after deleting a bunch of pods and scaling down to zero, I ended up with a few leftovers:

I think the replacement pods are being layered on top of other active pod positions, instead of being rescheduled into available empty slots (which may be leaving orphaned pod statuses on the board after the scale down)

[detiber]

Yes, I saw a similar thing when Burr and I presented at DevNexus.

[hodrigohamalho]

it's ok on OSE 3.1.6 :)

[ryanj]

This seems to be working just fine in the Origin v1.1.4 VM

[ryanj]

Not working in my current OSE environment :(

Workaround: Instead of showing autorecovery on the hexboard, show the OpenShift web console instead. It looks good there using both of the available view modes on your project's Overview page (list, topographical)