search

coreos / tectonic-forum

Apache License 2.0
30 stars 9 forks source link

Upgrade issue from v1.6.2 - webproxy? #252

Open esselfour opened 6 years ago

esselfour commented 6 years ago

Issue Report Template

Tectonic Version

v1.6.2

Environment

Bare Metal 3 nodes - 2 workers behind a proxy

What hardware/cloud provider/hypervisor is being used with Tectonic? vmware

Expected Behavior

That the upgraded versions are found by console and can proceed.

Actual Behavior

The console states it is up to date

Reproduction Steps

  1. Check for new updates
  2. No new updates found

Other Information

Sorry for the lack of detail, I'm new to all of this. Linux admin who has been passed this environment as the last guy has left the organisation!

I'm trying to update the cluster but it appears it cannot connect.

We are behind a web proxy and see this..

kubectl -n tectonic-system logs tectonic-channel-operator-1363 898206-5mfdp :... ... E0109 14:40:49.209919 1 main.go:125] Failed to get TectonicVersion from Co reUpdate: Post https://tectonic.update.core-os.net/v1/update: dial tcp 34.192.32 .119:443: getsockopt: connection refused

Is this because of the proxy? The cluster otherwise can access the world outside.

Failing that, can update this manually?

Many thanks.

Feature Request

Environment

What hardware/cloud provider/hypervisor is being used with Tectonic?

Desired Feature

Other Information

esselfour commented 6 years ago

To add, the web proxy environment variables have been set already.

kbrwn commented 6 years ago

@esselfour how did you set up the proxy? Did you follow any particular documentation or a guide?

You might try running another container in the tectonic-system namespace to ensure it is properly inheriting proxy values from the host: 

$ kubectl -n tectonic-system run -it fedora-tool --image=fedora  -- curl https://tectonic.update.core-os.net/v1/update
CoreOS Update Service. Expected a POST. For details see: http://coreos.com
esselfour commented 6 years ago

@kbrwn

Thanks for getting back to me.

The webproxy is just a standard web proxy we supply for internal machines to access the outside world for updates etc when needed.

I tried running the line you supplied, the deployment is visible but the log on the pod states this:

curl: (7) Failed to connect to tectonic.update.core-os.net port 443: Connection refused

The tectonics nodes all have the proxy details using various methods that are also found on this forum, for example the 10-default-env.conf under systemd and /etc/environment.

Looked at issue #235 and tried editing the deployment ' tectonic-channel-operator' with the proxy details. This made a difference but not a positive one - the check for updates button just sat there hanging looking like something was happening but nothing did!

Is there anything I could have missed somewhere?

Thanks in advance.

esselfour commented 6 years ago

Bit of an update, after playing around with the 'tectonic-channel-operator' I also managed to get around my issue by injecting the web-proxy env values into the deployment as noted in #235 . This went OK for 2 versions update iterations till trying to update to 1.6.10-tectonic.3 where the channel operator is updated. I see exactly the same issue as in #223 and known issues with supporting proxies as shown in https://github.com/coreos/tectonic-installer/issues/38

The update installs, terminates the pod and then loops back around to install again and loses the environment vars input previously..

Does anyone know how I could get around this?