In containerd mode it looks like when you request a force kill of the last container in a VM via docker kill -s KILL somecontainer and the VM shutdown timeout is not met, runv will force-kill the VM in the container but then does not update docker or its internal state properly to indicate that it has done so.
The result is that the docker kill command never returns (until it HTTP timeouts) and subsequent docker ps commands erroneously report the container as still running despite the fact the VM has been force-killed.
In containerd mode it looks like when you request a force kill of the last container in a VM via
docker kill -s KILL somecontainer
and the VM shutdown timeout is not met,runv
will force-kill the VM in the container but then does not update docker or its internal state properly to indicate that it has done so.The result is that the
docker kill
command never returns (until it HTTP timeouts) and subsequentdocker ps
commands erroneously report the container as still running despite the fact the VM has been force-killed.