Without this buffering, the compressor was outputting 64 bytes at a
time to the HTTP stream, which was resulting in absurdly small chunk
sizes and a lot of extra overhead. The buffering restores the chunk size
to 32768 bytes, which matches the behavior with 1.8.2.
Times pushing to a local registry:
1.8.2: 0m18.934s
master: 0m20.564s
master+this commit: 0m17.593s
Fixes: #17038
Signed-off-by: Aaron Lehmann <aaron.lehmann@docker.com>
- Do not execute the template directly in the cli outout, go is not atomic
in this operation and can send bytes before failing the execution.
- Fail after evaluating a raw interface if the typed execution also
failed, assuming there is a template parsing error.
Signed-off-by: David Calavera <david.calavera@gmail.com>
This carries fixes for
- Internal racy /etc/hosts updates within container during SD
- Renable SD service record watch after cluster-store restarts
- Fix to allow remote IPAM driver to return no IP if the user prefers
- Fix to allow --fixed-cidr and --bip to be in same range
Signed-off-by: Madhu Venugopal <madhu@docker.com>
PR #15493 removed compression of layers when pushing them to a V2
registry. This this makes layer uploads larger than they should be.
This commit restores the compression. It uses an io.Pipe to turn the
gzip compressor output Writer into a Reader, so the ReadFrom method can
be used on the BlobWriter (which is very important for avoiding many
PATCH requests per layer).
Fixes#17209Fixes#17038
Signed-off-by: Aaron Lehmann <aaron.lehmann@docker.com>
Also remove unused func `newTempFile` and prevent a possible deadlock
between pull_v2 `attemptIDReuse` and graph `register`
Signed-off-by: Antonio Murdaca <runcom@redhat.com>
Lookup the graph parent reference to detect a builder cache miss before
looping the whole graph image index to build a parent-children tree.
Signed-off-by: Antonio Murdaca <runcom@redhat.com>
Delegate validation tasks to the volume drivers. It's up to them
to decide whether a name is valid or not.
Restrict volume names for the local driver to prevent creating
mount points outside docker's volumes directory.
Signed-off-by: David Calavera <david.calavera@gmail.com>
Since #15862, containers fail to start when started with --ipc=host if
/dev/mqueue is not present. This change causes docker to create
container-local mounts for --ipc=host containers as well as in the
default case.
Signed-off-by: Burke Libbey <burke.libbey@shopify.com>
To be consistent with other inspect command (on container and images),
add the possiblity to pass multiple network to the network inspect
commands.
`docker network inspect host bridge none` is possible now.
Signed-off-by: Vincent Demeester <vincent@sbr.pm>
After v1.8.3 layer checksum is used for image ID
validation. Rewriting the checksums on push would
mean that next pulls will get different image IDs
and pulls may fail if its detected that same
manifest digest can now point to new image ID.
Fixes#17178
Signed-off-by: Tonis Tiigi <tonistiigi@gmail.com>
cleanupDeleted() takes devices.Lock() but does not drop it if there are
no deleted devices. Hence docker deadlocks if one is using deferred
device deletion feature. (--storage-opt dm.use_deferred_deletion=true).
Fix it. Drop the lock before returning.
Also added a unit test case to make sure in future this can be easily
detected if somebody changes the function.
Signed-off-by: Vivek Goyal <vgoyal@redhat.com>
This will prevent the api from trying to serve network requests in
systems where libnetwork is not enabled, returning 404 responses in any
case.
Signed-off-by: David Calavera <david.calavera@gmail.com>
Aaron Lehmann