telepresence should cleanup/kill its own inactive an expired pods

[sokoow]

Hi All,

Telepresence currently has problems cleaning up pods and deployments after itself. Here's a snapshot after couple of days of work:

pods:

telepresence-1547700365-423302-47147-6875bc4cc8-bgz5g   1/1     Running     0          16d
telepresence-1547702626-319927-47984-6fb967cb5-nx4nb    1/1     Running     0          16d
telepresence-1547706680-838876-49122-77b868bf7d-bcqg9   1/1     Running     0          16d
telepresence-1547712456-517427-54707-b5776bfdb-vj4bl    1/1     Running     0          16d
telepresence-1547712471-276171-54786-58f679576-5kw8k    1/1     Running     0          16d
telepresence-1547713700-630524-55675-775d5b8bf8-tqgx7   1/1     Running     0          16d
telepresence-55dffdcfc9-7967h                           1/1     Running     0          30d

deployments:

telepresence                           1/1     1            1           30d
telepresence-1547700365-423302-47147   1/1     1            1           20d
telepresence-1547702626-319927-47984   1/1     1            1           20d
telepresence-1547706680-838876-49122   1/1     1            1           20d
telepresence-1547712456-517427-54707   1/1     1            1           20d
telepresence-1547712471-276171-54786   1/1     1            1           20d
telepresence-1547713700-630524-55675   1/1     1            1           20d

Not ideal.

[ark3]

Agreed! Sorry about that. See also #260 (comment).

[sokoow]

Hey no worries, it's a great idea and great project, just some corners to smooth out

[ark3]

Out of curiosity, did those deployments get created using the (default) new deployment operation? Were any services left behind? I have an idea that I'd like to explore...

[sokoow]

I think devs just left these behind, closed their laptops and telepresence lost their connection to the cluster. Afterwards, when they opened their laptops again, connections expired, leaving the proxy hanging.

[maelvls]

We also have the issue of 'interrupted' telepresence connections leaving many dangling deployments & pods.

Just wondering, is there an easy way (with kubectl) to cleanup all telepresence deployments that have been created more than 1d ago? Since a telepresence session is often short, any deployments older than 1 day are probably leftovers 😁

For now, I use

kubectl delete deployments -l telepresence

but it might as well kill 'in-use' deployments

[PPACI]

kubectl get pods -A -l telepresence -o json | jq -r '.items[] |select((.metadata.creationTimestamp | now - fromdateiso8601) > 86400) | @sh "kubectl delete --grace-period=1 -n \(.metadata.namespace) pod/\(.metadata.name)"' | xargs
For people coming here from Google searching for a way to clean expired pods, this will delete all telepresence older than 1 day. Note that this require JQ to be installed. I'm sure there is a way to do it with only kubectl, but I was a bit lazy 🙃.

[donnyyung]

I believe this is no longer an issue in Telepresence 2 since you can use the telepresence uninstall command to uninstall agents + the traffic-manager. Here are the docs on how to install Telepresence (https://www.telepresence.io/docs/latest/install/), please re-open if you still see this issue in our latest version!

[QuinnBast]

The uninstall command posted above just tries to install the traffic-manager again.

$ telepresence uninstall --everything
Launching Telepresence Root Daemon
Launching Telepresence User Daemon
...