kubed-sh left a bunch of pods running on my cluster after exit?

[plombardi89]

Wondering if this is normal behavior. I started kubed-sh then typed exit once launched and checked what was running and all the kubed-sh prepull-xyz pods were still running. I expected them to be terminated after I was done ...

default                    prepullbin-2gbk8                                      1/1       Running   0          7m
default                    prepullbin-75dcx                                      1/1       Running   0          7m
default                    prepullbin-c8234                                      1/1       Running   0          7m
default                    prepullbin-csnr8                                      1/1       Running   0          7m
default                    prepullbin-k31n3                                      1/1       Running   0          7m
default                    prepullbin-rdpsz                                      1/1       Running   0          7m
default                    prepulljs-56vwz                                       1/1       Running   0          7m
default                    prepulljs-6mb07                                       1/1       Running   0          7m
default                    prepulljs-8xln7                                       1/1       Running   0          7m
default                    prepulljs-dj6zz                                       1/1       Running   0          7m
default                    prepulljs-npb43                                       1/1       Running   0          7m
default                    prepulljs-nq5dr                                       1/1       Running   0          7m
default                    prepullpy-7mdvn                                       1/1       Running   0          7m
default                    prepullpy-8x9tx                                       1/1       Running   0          7m
default                    prepullpy-9dshc                                       1/1       Running   0          7m
default                    prepullpy-b6lv4                                       1/1       Running   0          7m
default                    prepullpy-dg440                                       1/1       Running   0          7m
default                    prepullpy-plk0f                                       1/1       Running   0          7m
default                    prepullrb-2p5x2                                       1/1       Running   0          7m
default                    prepullrb-3nhd8                                       1/1       Running   0          7m
default                    prepullrb-85hpj                                       1/1       Running   0          7m
default                    prepullrb-8qbh5                                       1/1       Running   0          7m
default                    prepullrb-gswjx                                       1/1       Running   0          7m
default                    prepullrb-lnjc9                                       1/1       Running   0          7m

[mhausenblas]

Thanks for trying out kubed-sh @plombardi89. This is indeed a bug (or better say: lack of proper garbage collection). You can get rid of it by issuing the following command:

kubectl delete ds --selector=gen=kubed-sh,scope=pre-flight 

If and when these DaemonSets are not around you will notice that kubed-sh on start-up will try to create them. You can disable this behaviour as described in the config section by launching kubed-sh with KUBEDSH_NOPREPULL=true.

Hope this helps?

[plombardi89]

Your workaround cleaned it up. Thanks :)

[mhausenblas]

I'll have to think longer about how to handle this properly. Given that without pre-pulling the images the first launch of a program will fail, I don't see a good way right now (other than the user disables it explicitly using KUBEDSH_NOPREPULL=true. Will revisit this issue in a later release.