When the daemon is going down trigger immediate
garbage collection of libnetwork resources deleted
like namespace path since there will be no way to
remove them when the daemon restarts.
Signed-off-by: Jana Radhakrishnan <mrjana@docker.com>
Upstream-commit: c68e7f96f9636a9b2ab0c2c0dbf753161fa73fc2
Component: engine
Also noticed potential hang when only stdout or stderr are used with
follow=1
Signed-off-by: Brian Goff <cpuguy83@gmail.com>
Upstream-commit: 0c84604f5458bc38b793e5bcdf86624eef3e3184
Component: engine
container.config.NetworkDisabled is set for both daemon's
DisableNetwork and --networking=false case. Hence using
this flag instead to fix#13725.
There is an existing integration-test to catch this issue,
but it is working for the wrong reasons.
Signed-off-by: Madhu Venugopal <madhu@docker.com>
Upstream-commit: 83208a531d1824ea7b0a55f9419e915e446c780a
Component: engine
vols.VolumesRW has been initialized so it can't be nil. Furthermore
it's ok to read a nil map.
Signed-off-by: Zefan Li <lizefan@huawei.com>
Upstream-commit: 8b4c0decfc6a3ef748a70c2f53d1351d5aa8ab64
Component: engine
Move some calls to container.LogEvent down lower so that there's
less of a chance of them being missed. Also add a few more events
that appear to have been missed.
Added testcases for new events: commit, copy, resize, attach, rename, top
Signed-off-by: Doug Davis <dug@us.ibm.com>
Upstream-commit: 8232312c1e705753d3db82dca3d9bb23e59c3b52
Component: engine
Merge user specified devices correctly with default devices.
Otherwise the user specified devices end up without permissions.
Signed-off-by: David R. Jenni <david.r.jenni@gmail.com>
Upstream-commit: c913c9921b7caa11e8500d2ebb4fa97303c876be
Component: engine
When using a scanner, log lines over 64K will crash the Copier with
bufio.ErrTooLong. Subsequently, the ioutils.bufReader will grow without
bound as the logs are no longer being flushed to disk.
Signed-off-by: Burke Libbey <burke.libbey@shopify.com>
Upstream-commit: f779cfc5d8b8f0686c3849a3bd4ab0480675180d
Component: engine
daemon.Diff already implements mounting for naivegraphdriver and
aufs which does diffing on its owns does not need the container to be mounted.
So new filesystem driver should mount filesystems on their own if it is needed
to implement Diff(). This issue was reported by @kvasdopil while working on a
freebsd port, because freebsd does not allow mount an already mounted
filesystem. Also it saves some cycles for other operating systems as well.
Signed-off-by: Jörg Thalheim <joerg@higgsboson.tk>
Upstream-commit: 6473b0f127c63e8a45b2b456d69e3de03273705c
Component: engine
I ran into a situation where I was trying:
`docker rmi busybox`
and it kept failing saying:
`could not find image: Prefix can't be empty`
While I have no idea how I got into this situation, it turns out this is
error message is from `daemon.canDeleteImage()`. In that func we loop over
all containers checking to see if they're using the image we're trying to
delete. In my case though, I had a container with no ImageID. So the code
would die tryig to find that image (hence the "Prefix can't be empty" err).
This would stop all processing despite the fact that the container we're
checking had nothing to do with 'busybox'.
My change logs the bad situation in the logs and then skips that container.
There's no reason to fail all `docker rmi ...` calls just because of one
bad container.
Will continue to try to figure out how I got a container w/o an ImageID
but as of now I have no idea, I didn't do anything but normal docker cli
commands.
Signed-off-by: Doug Davis <dug@us.ibm.com>
Upstream-commit: 71a499022904e1da074ded3e1ed874c1b75ccf23
Component: engine