When the daemon is started, it looks at all the volumes and checks to
see if any of them have mount options persisted to disk, and loads them
from disk if it does.
In some cases a volume will be created with an empty map causing the
options file to be persisted and volume options set to a non-nil value
on daemon restart... this causes problems later when the driver checks
for a non-nil value to determine if it should try and mount with the
persisted volume options.
Ensures 2 things:
1. Instead of only checking nilness for the opts map, use `len` to make
sure it is not an empty map, which we don't really need to persit.
2. An empty (or nulled) opts.json will not inadvertnatly set volume
options on daemon restart.
Signed-off-by: Brian Goff <cpuguy83@gmail.com>
(cherry picked from commit 246d1eb58e)
Signed-off-by: Tibor Vass <tibor@docker.com>
This example was added in b0b2f979c7,
but got lost during splitting up the cli.md docs into separate files;
561bfb268d
This restores the missing example.
Signed-off-by: Sebastiaan van Stijn <github@gone.nl>
(cherry picked from commit 8d5818a693)
Signed-off-by: Sebastiaan van Stijn <github@gone.nl>
Commit 067b510dfc move the
sysctls feature to the right API version, but
missed this one.
Signed-off-by: Sebastiaan van Stijn <github@gone.nl>
(cherry picked from commit 7cdd693d5d)
Signed-off-by: Sebastiaan van Stijn <github@gone.nl>
The service create/update endpoints support
passing registry-authentication information,
but this was not documented.
This adds the missing information about passing
a X-Registry-Auth header.
Signed-off-by: Sebastiaan van Stijn <github@gone.nl>
(cherry picked from commit bde35ef18a)
Signed-off-by: Sebastiaan van Stijn <github@gone.nl>
The export command operates on containers, not images, so it should be listed under the container commands, not the image commands.
Signed-off-by: David Lechner <david@lechnology.com>
(cherry picked from commit 43c8cc7bb4)
Signed-off-by: Sebastiaan van Stijn <github@gone.nl>
Volumes and other content created under a bind mount should be
recursively propagated using rshared, not shared. This could be
the reason for EBUSY during removal. Override options with rbind,
rshared and see if CI errors are fixed.
May fix#25511
Signed-off-by: Anusha Ragunathan <anusha@docker.com>
(cherry picked from commit e58028d078)
Signed-off-by: Tibor Vass <tibor@docker.com>
- Fixes a SNAT issue in loadbalancer when multiple networks are involved
- Fixes an issue with SRV query forwarding in DNS
- Fixes a map race
Signed-off-by: Jana Radhakrishnan <mrjana@docker.com>
(cherry picked from commit 56c9822695)
Signed-off-by: Tibor Vass <tibor@docker.com>
This is an attempt to fix the flaky test of TestSwarmNodeTaskListFilter in 25029.
Basically this fix adds a check to wait until 3 containers has already up,
before processing `node tasks ...`.
This might fix 25029.
Signed-off-by: Yong Tang <yong.tang.github@outlook.com>
(cherry picked from commit 63c0366bc9)
Signed-off-by: Tibor Vass <tibor@docker.com>
These flags were not supported (daemon returns an error), and it was an
oversight. They were not present in completion scripts.
Signed-off-by: Tibor Vass <tibor@docker.com>
(cherry picked from commit 3a7e90c743)
Signed-off-by: Tibor Vass <tibor@docker.com>
/proc/timer_list seems to leak information about the host. Here is
an example from a busybox container running on docker+kubernetes.
# cat /proc/timer_list | grep -i -e kube
<ffff8800b8cc3db0>, hrtimer_wakeup, S:01, futex_wait_queue_me, kubelet/2497
<ffff880129ac3db0>, hrtimer_wakeup, S:01, futex_wait_queue_me, kube-proxy/3478
<ffff8800b1b77db0>, hrtimer_wakeup, S:01, futex_wait_queue_me, kube-proxy/3470
<ffff8800bb6abdb0>, hrtimer_wakeup, S:01, futex_wait_queue_me, kubelet/2499
Signed-Off-By: Davanum Srinivas <davanum@gmail.com>
Signed-off-by: Davanum Srinivas <davanum@gmail.com>
(cherry picked from commit 03bd00b68f)
Signed-off-by: Tibor Vass <tibor@docker.com>
