Endpoint disconnect ("We've been having trouble communicating with Cloud Foundry")

[svollath]

Frontent Deployment type

• Cloud Foundry Application (cf push)
• Kubernetes, using a helm chart
• Docker, using docker compose
• Docker, single container deploying all components
• npm run start
• Other (please specify below)

Backend (portal-proxy) Deployment type

• Cloud Foundry Application (cf push)
• Kubernetes, using a helm chart
• Docker, using docker compose
• Docker, single container deploying all components
• Other (please specify below)

Stratos 2.0.0-rc1 with pre SCF-2.11.0-rc1 on CaaSPv3-GM

Expected behaviour

The UI should either keep the session active until the user logs out, or redirect to the login page after a default session timeout.

Actual behaviour

When the UI isn't touched for a while the following error will be shown:

"We've been having trouble communicating with Cloud Foundry - You may be seeing out-of-date information"

Steps to reproduce the behavior

Log in to Stratos as admin, and leave the browser window open (idle) for a while (~10 min.). Then try to reuse the UI.

Log output covering before error and any error statements

"We've been having trouble communicating with Cloud Foundry - You may be seeing out-of-date information"

When I "Logout" and "Login" it works again, but will also loose communication after a while again.

Looks like the session times out by default, but the user isn't redirected to the login page. Instead the error message is shown.

[richard-cox]

This sounds like a bug. We should show an 'Are you there?' prompt which if left should automatically log the user out. Does this not show at all? When you see this behaviour are you on an application page or another part of Stratos?

[svollath]

JFYI: Now, the same error occured while I was constantly using the UI, so it's maybe not related to a default session timeout.

[richard-cox]

Phew, ok. Are there any interesting statements in the browser's console? We'll show the error bar if a request to a specific CF fails. There should be more information in the request's response. Note - Requests can go out to multiple CF's, so the response for a specific CF response will be nested under it's guid.

[svollath]

There's one 401 for "verify", which is when the URL is given the first time, then one is redirected to the login page.
Now I reproduced the issue again, and then there's:

"WebSocket connection to 'wss://10.160.67.203.xip.io:8443/pp/v1/0bfc15ad-6595-43c1-9676-700fb5c02ec0/firehose' failed: HTTP Authentication failed; no valid credentials available" for "main".

But that happens just when I try to access "CloudFoundry"->"Firehose".
No error was logged, when it happened, or when I just switch between different menus.

Beside that there are only dozens of messages, saying

"[Violation] Added non-passive event listener to a scroll-blocking 'mousewheel' event. Consider marking event handler as 'passive' to make the page more responsive. See https://www.chromestatus.com/feature/5745543795965952"

which don't seem related.

[svollath]

Logs from the proxy container, when error appears:

ERRO[Fri Jun 29 13:34:17 UTC 2018] Unable to save a CNSI Token: Unable to UPDATE CNSI token: no rows were updated 
INFO[Fri Jun 29 13:34:17 UTC 2018] Couldn't save new token: Unable to save a CNSI Token: Unable to UPDATE CNSI token: no rows were updated 
WARN[Fri Jun 29 13:34:17 UTC 2018] Passthrough response: URL: https://api.10.160.67.203.xip.io/v2/apps/c7ec9831-3bc5-42e5-9c3c-286e8ee8e954/stats?order-direction=desc&order-direction-field=index&results-per-page=5&page=1, Status Code: 500, Status: Error proxing request, Content Type: Unknown, Length: -1 
WARN[Fri Jun 29 13:34:17 UTC 2018] Couldn't refresh token for CNSI with GUID 0bfc15ad-6595-43c1-9676-700fb5c02ec0 
Request: [2018-06-29T13:34:17Z] Remote-IP:"172.16.4.0" Method:"GET" Path:"/v1/proxy/v2/apps/c7ec9831-3bc5-42e5-9c3c-286e8ee8e954/stats" Status:200 Latency:2.575411902s Bytes-In:0 Bytes-Out:200```

[irfanhabib]

Some more detail from @svollath

yes, it exactly happend now with RC2 - the deployment succeeded 20 min. ago and I was logged in for maybe 10-15 min. - I just let "Cloudfoundry"-"Firehose" run, and the switched to "Applications" after ~10min.. It could be that it's triggered by switching to another menu after a while.

Irfan Habib @IHabib 2:58 PM
ok I'll keep that in mind when testing
thanks

Sebastian Vollath @svollath 3:02 PM
... and btw. when it's in that state (Communication Error) - then, after another while "Are you still there" and redirect to the login page will both work. So in fact my initial thought is wrong. So feel free to change the title of that issue ...

Irfan Habib @IHabib 3:03 PM
So basically...
1. The redirect works when the `Are you still there` message pops up
2. What doesn't work is: there is an endpoint error (red bar) when you perform some operations after soem time?
is this correct?

Sebastian Vollath @svollath 3:06 PM
Yes, that's what I got from Friday - when I just left the UI from reproducing, I saw a "Are you still there" dialog, as well as it was back to the login page - but I mean that must have happened more than an hour later ...

[svollath]

Like mentioned in a comment before already the relation to session timeouts was wrong. I reproduced it with 2.0.0-rc2 as well, and even when it's disconnected, the session will time out and redirect as expected.

So the issue is about loosing communication to cloud foundry under UI usage - the proxy container logs "Couldn't refresh token for CNSI".