Add a mutable state to pods, and database backend sutable for
modifying and updating said state.
Signed-off-by: Matthew Heon <matthew.heon@gmail.com>
Closes: #784
Approved by: rhatdan
Instead of manually calling the individual functions that cleanup
uses to tear down a container's resources, just call the cleanup
function to make sure that cleanup only needs to happen in one
place.
Signed-off-by: Matthew Heon <matthew.heon@gmail.com>
Closes: #790
Approved by: rhatdan
When we're given a numeric --user value, default to GID 0 if the numeric
ID doesn't correspond to a user entry in /etc/passwd that can provide us
with the user's primary group ID.
Make sure that GetAdditionalGroupsForUser() returns wrapped errors.
Also test various user:group forms.
Signed-off-by: Nalin Dahyabhai <nalin@redhat.com>
Closes: #728
Approved by: mheon
If the user uses the image ID when saving to either docker-archive
or oci-archive, then do not save a reference in the manifest/index.json.
If the user chooses to push without an image reference, i.e <transport>:<path>
it should be valid and succeed.
Signed-off-by: umohnani8 <umohnani@redhat.com>
Closes: #782
Approved by: rhatdan
Allow containers that no longer exist in storage to be evicted
from the state instead of erroring.
Signed-off-by: Matthew Heon <matthew.heon@gmail.com>
Closes: #764
Approved by: rhatdan
And add an argument to WithHooksDir to set it.
If the hook dir doesn't exist, the new hooks package considers that a
fatal error. When a podman caller sets
--hooks-dir-path=/some/typoed/directory, a fatal error is more helpful
than silently not loading any hooks. However, callers who call podman
without setting --hooks-dir-path may not need hooks at all. We don't
want to pester those callers with not-exist errors. With this commit,
we:
* Assume the caller knows what they're doing if they set
--hooks-dir-path and set HooksDirNotExistFatal.
* If the caller does not explicitly set --hooks-dir-path, assume they
won't mind if the hook directory is missing and set
HooksDirNotExistFatal false.
We also considered checking for the directory's existence in the code
calling WithHooksDir or from within WithHooksDir, but checks there
would race with the underlying ioutil.ReadDir in the hooks package.
By pushing the warn/error decision down into libpod's implementation,
we avoid a racy "do we expect this to work once libpod gets to it?"
pre-check.
I've also added a check to error if WithHooksDir is called with an
empty-string argument, because we haven't defined the semantics of
that (is it clearing a previous value? Is it effectively the same as
the current directory?). I agree with Matthew that a separate
WithNoHooks, or a *string argument to WithHooks, or some such would be
a better API for clearing previous values [1]. But for now, I'm just
erroring out to fail early for callers who might otherwise be
surprised that libpod ignores empty-string HooksDir.
[1]: https://github.com/projectatomic/libpod/pull/686#issuecomment-385119370
Signed-off-by: W. Trevor King <wking@tremily.us>
Closes: #686
Approved by: mheon
This shifts the matching logic out of libpod/container_internal and
into the hook package, where we can reuse it after vendoring into
CRI-O. It also adds unit tests with almost-complete coverage. Now
libpod is even more isolated from the hook internals, which makes it
fairly straightforward to bump the hook config file to 1.0.0. I've
dubbed the old format 0.1.0, although it doesn't specify an explicit
version. Motivation for some of my changes with 1.0.0:
* Add an explicit version field. This will make any future JSON
structure migrations more straightforward by avoiding the need for
version-guessing heuristics.
* Collect the matching properties in a new When sub-structure. This
makes the root Hook structure easier to understand, because you
don't have to read over all the matching properties when wrapping
your head around Hook.
* Replace the old 'hook' and 'arguments' with a direct embedding of
the runtime-spec's hook structure. This provides access to
additional upstream properties (args[0], env, and timeout) and
avoids the complication of a CRI-O-specific analog structure.
* Add a 'when.always' property. You can usually accomplish this
effect in another way (e.g. when.commands = [".*"]), but having a
boolean explicitly for this use-case makes for easier reading and
writing.
* Replace the previous annotations array with an annotations map. The
0.1.0 approach matched only the values regardless of key, and that
seems unreliable.
* Replace 'cmds' with 'when.commands', because while there are a few
ways to abbreviate "commands", there's only one way to write it out
in full ;). This gives folks one less thing to remember when
writing hook JSON.
* Replace the old "inject if any specified condition matches" with
"inject if all specified conditions match". This allows for more
precise targeting. Users that need more generous targeting can
recover the previous behavior by creating a separate 1.0.0 hook file
for each specified 0.1.0 condition.
I've added doc-compat support for the various pluralizations of the
0.1.0 properties. Previously, the docs and code were not in
agreement. More on this particular facet in [1].
I've updated the docs to point out that the annotations being matched
are the OCI config annotations. This differs from CRI-O, where the
annotations used are the Kubernetes-supplied annotations [2,3]. For
example, io.kubernetes.cri-o.Volumes [4] is part of CRI-O's runtime
config annotations [5], but not part of the Kubernetes-supplied
annotations CRI-O uses for matching hooks.
The Monitor method supports the CRI-O use-case [6]. podman doesn't
need it directly, but CRI-O will need it when we vendor this package
there.
I've used nvidia-container-runtime-hook for the annotation examples
because Dan mentioned the Nvidia folks as the motivation behind
annotation matching. The environment variables are documented in [7].
The 0.1.0 hook config, which does not allow for environment variables,
only works because runc currently leaks the host environment into the
hooks [8]. I haven't been able to find documentation for their usual
annotation trigger or hook-install path, so I'm just guessing there.
[1]: https://github.com/kubernetes-incubator/cri-o/pull/1235
[2]: https://github.com/kubernetes-incubator/cri-o/blob/v1.10.0/server/container_create.go#L760
[3]: https://github.com/kubernetes-incubator/cri-o/blob/v1.10.0/server/container_create.go#L772
[4]: https://github.com/kubernetes-incubator/cri-o/blob/v1.10.0/pkg/annotations/annotations.go#L97-L98
[5]: https://github.com/kubernetes-incubator/cri-o/blob/v1.10.0/server/container_create.go#L830-L834
[6]: https://github.com/kubernetes-incubator/cri-o/pull/1345/
[7]: https://github.com/NVIDIA/nvidia-container-runtime/tree/v1.3.0-1#environment-variables-oci-spec
[8]: https://github.com/opencontainers/runc/pull/1738
Signed-off-by: W. Trevor King <wking@tremily.us>
Closes: #686
Approved by: mheon
Like podman pull, when you push an image, podman should check
if the registry is listed as insecure and if so, it should
--tls-verify=false unless the user overrides this.
Signed-off-by: baude <bbaude@redhat.com>
Closes: #738
Approved by: mheon
the error was:
libpod/container.go:219::error: struct field tag `json:"groups, omitempty"` not compatible with reflect.StructTag.Get: suspicious space in struct tag value (vet)
Signed-off-by: Giuseppe Scrivano <gscrivan@redhat.com>
Closes: #729
Approved by: baude
This prevents you from cleaning up the container database, if
some how runc and friends db gets screwed up.
Signed-off-by: Daniel J Walsh <dwalsh@redhat.com>
Closes: #725
Approved by: mheon
so that the OCI runtime creates the network namespace from the correct
userNS.
Signed-off-by: Giuseppe Scrivano <gscrivan@redhat.com>
Signed-off-by: Daniel J Walsh <dwalsh@redhat.com>
Closes: #690
Approved by: mheon
when pulling from a secure registry that is documented in registries.conf, we
should be able to pull without tls-verify=false
Signed-off-by: baude <bbaude@redhat.com>
Closes: #718
Approved by: rhatdan
This allows us to restart containers that have never been started
without error. This makes RestartWithTimeout work with running,
stopped, and created containers.
Signed-off-by: Matthew Heon <matthew.heon@gmail.com>
Closes: #719
Approved by: rhatdan
first pass at adding in the container related endpoints/methods for the libpod
backend. Couple of important notes:
* endpoints that can use a console are not going to be done until we have "remote" console
* several of the container methods should probably be able to stream as opposed to a one-off return
Signed-off-by: baude <bbaude@redhat.com>
Closes: #708
Approved by: baude
We need these for commit, and they cannot be properly deduced
from just the OCI spec, so save them in the database so we can
retrieve them for commit.
Signed-off-by: Matthew Heon <matthew.heon@gmail.com>
Closes: #700
Approved by: rhatdan
This allows us to accurately trigger OCI hooks that trigger on
the presence of volume mounts.
Signed-off-by: Matthew Heon <matthew.heon@gmail.com>
Closes: #700
Approved by: rhatdan
when performing a container commit, we should not add the default list of volumes
for a container to the resulting image. it will cause the resulting image to crash
when run subsequently.
Signed-off-by: baude <bbaude@redhat.com>
Closes: #699
Approved by: mheon
Vendor in buildah and use as much of commit and bug as possible for podman
build and commit.
Resolves#586
Signed-off-by: baude <bbaude@redhat.com>
Closes: #681
Approved by: mheon
Made a mistake in my earlier patch. I though that if you add an empty string
to an array, the length of the array would still be 0...
Realised this when vendoring the secrets pkg into cri-o.
Signed-off-by: umohnani8 <umohnani@redhat.com>
Closes: #685
Approved by: mheon
implement varlink image functions for working with libpod with the exception of a
couple due to incompletions on the libpod side of things (build).
also, created a first pass at a libpodpy package which will stand as a client to
working with libpod's varlink methods using python.
Signed-off-by: baude <bbaude@redhat.com>
Closes: #669
Approved by: baude
Instead of execing out to the host's IP, use the IP address we
got back from CNI to populate Inspect's IP address information.
Signed-off-by: Matthew Heon <mheon@redhat.com>
Closes: #680
Approved by: umohnani8
The hidden flag is used to override the path of the default mounts file
for testing purposes.
Also modified the secrets pkg to allow for this override to happen.
Signed-off-by: umohnani8 <umohnani@redhat.com>
Closes: #678
Approved by: mheon
When a user pulls an image using a transport, like docker-daemon, we try to lookup
the new image in storage by the input name after the pull. Because the input name
has a transport (different than local storage), that lookup would fail.
Signed-off-by: baude <bbaude@redhat.com>
Closes: #644
Approved by: rhatdan
If an integer is passed into the --user flag, i.e --user=1234
don't look up the user in /etc/passwd, just assign the integer as the uid.
Signed-off-by: umohnani8 <umohnani@redhat.com>
Closes: #652
Approved by: mheon
If the host is in FIPS mode and /etc/system-fips exists
/run/secrets/system-fips is created in the container so that
the container can run in FIPS mode as well.
Signed-off-by: umohnani8 <umohnani@redhat.com>
Made necessary changes to functions to include contex.Context wherever needed
Signed-off-by: umohnani8 <umohnani@redhat.com>
Closes: #640
Approved by: baude
In the case where podman needs to pull an image, if that registry that the image
resides on is known to be insesure (as defined in /etc/containers/registries.conf),
tls-verify should be altered on the fly.
Signed-off-by: baude <bbaude@redhat.com>
Closes: #626
Approved by: mheon
In the case where you have an image local, if the the user runs
podman pull, we should always attempt to pull an updated image.
Added a forceRemote bool to New (image) so we can differentiate
between "pull" or run because the actions differ. Run does not
need to pull the latest -- only run.
Signed-off-by: baude <bbaude@redhat.com>
Closes: #618
Approved by: baude
Comparing Go interfaces, like io.Reader, to nil does not work. As
such, we need to include a bool with each stream telling whether
to attach to it.
Signed-off-by: Matthew Heon <matthew.heon@gmail.com>
Closes: #608
Approved by: baude
This allows us to attach to attach to just stdout or stderr or
stdin, or any combination of these.
Signed-off-by: Matthew Heon <matthew.heon@gmail.com>
Closes: #608
Approved by: baude
Until we can handle running containers which use UID/GID mappings, make
sure that we always create containers that use the host mappings.
Signed-off-by: Nalin Dahyabhai <nalin@redhat.com>
Closes: #609
Approved by: baude
This solves a nasty locking issue with getting the path of
namespaces for dependencies
Signed-off-by: Matthew Heon <matthew.heon@gmail.com>
Closes: #600
Approved by: rhatdan
When a container is transitioning from running to stopped and stats is runnings,
we should not break stats if we are unable to get stats for that container.
Resolves: #598
Signed-off-by: baude <bbaude@redhat.com>
Closes: #599
Approved by: mheon
--group-add
--blkio-weight-device
--device-read-bps
--device-write-bps
--device-read-iops
--device-write-iops
--group-add now supports group names as well as the gid associated with them.
All the --device flags work now with moderate changes to the code to support both
bps and iops.
Added tests for all the flags.
Signed-off-by: umohnani8 <umohnani@redhat.com>
Closes: #590
Approved by: mheon
If sending a signal fails, check if the container is alive. If it
is not, it probably stopped on its own before we could send the
signal, so don't error out.
Signed-off-by: Matthew Heon <matthew.heon@gmail.com>
Closes: #591
Approved by: rhatdan
In our ezrly development, we always allocated a tty when not -d. Now we should only allocated when the user asks for it.
Resolves: #573
Signed-off-by: baude <bbaude@redhat.com>
Closes: #574
Approved by: rhatdan
Instead of checking during init(), which could result in major
locking issues when used with pods, make our dependency checks in
the public API instead. This avoids doing them when we start pods
(where, because of the dependency graph, we can reasonably say
all dependencies are up before we start a container).
Signed-off-by: Matthew Heon <matthew.heon@gmail.com>
Closes: #577
Approved by: rhatdan
We don't use it directly, we aren't going to cache it in the DB,
and when we do use it (image volumes) we might well be in a
different process (podman create -> podman start). No point in
keeping it around.
Signed-off-by: Matthew Heon <matthew.heon@gmail.com>
Closes: #571
Approved by: rhatdan
We have other tools using containers/storage. They can delete our
containers in c/storage without us knowing. Try and handle this
better by warning instead of erroring when delete our storage and
it is already gone.
This does not handle cases where libpod thinks the container is
mounted, but it is not. This is harder to check for, because
c/storage Mount() and Unmount() take a layer, image, or container
and that complicates our "container no longer exists" question.
Further work is needed here.
Signed-off-by: Matthew Heon <matthew.heon@gmail.com>
Closes: #571
Approved by: rhatdan
We should allow users to pass in image ids with the sha256: prefix
for local images.
Resolves: #493
Signed-off-by: baude <bbaude@redhat.com>
Closes: #560
Approved by: baude
Adds support for mounting secrets especially on RHEL where the container
can use the host subsription to run yum
Signed-off-by: umohnani8 <umohnani@redhat.com>
Closes: #544
Approved by: rhatdan
Instead of looping to find containers with no dependencies,
maintain a map of them and remove entries as we add dependency
edges.
Signed-off-by: Matthew Heon <matthew.heon@gmail.com>
Closes: #557
Approved by: rhatdan
This fixes the situation where we fail to create a container when a name already exists.
Signed-off-by: Daniel J Walsh <dwalsh@redhat.com>
Closes: #517
Approved by: baude
so that it is possible to use systemd to automatically restart the
container:
[Service]
Type=forking
PIDFile=/run/awesome-service.pid
ExecStart=/usr/bin/podman run --conmon-pidfile=/run/awesome-service.pid --name awesome -d IMAGE /usr/bin/do-something
ExecStopPost=/usr/bin/podman rm awesome
Restart=always
Closes: https://github.com/projectatomic/libpod/issues/534
Signed-off-by: Giuseppe Scrivano <gscrivan@redhat.com>
Closes: #549
Approved by: rhatdan
Containers in pods cannot depend on containers outside of the
same pod. Make the reverse true as well - containers not in pods
cannot depend on containers in pods. This greatly simplifies our
dependency handling, as we can guarantee that removing a pod will
not encounter dependency issues.
Signed-off-by: Matthew Heon <matthew.heon@gmail.com>
Closes: #558
Approved by: rhatdan
podman parse and attach were using a very small portion of the kubernetes code
but using it caused a signficant increase in binary size.
Signed-off-by: baude <bbaude@redhat.com>
Closes: #559
Approved by: rhatdan
When no image is found, display a useful error message. Also, in imageToRef
protect against a nil image being passed.
Resolves: #553
Signed-off-by: baude <bbaude@redhat.com>
Closes: #555
Approved by: mheon
The SQLite DB backend has become an unmanageable nightmare. I
like having the option for DB work, but it's become an active
hindrance to further development, and it's definitely not in any
shape to be actively used.
Signed-off-by: Matthew Heon <matthew.heon@gmail.com>
Closes: #548
Approved by: baude
Add only when it's not already present.
Add a more specific version in podman spec generation
so we get 'container=podman' not 'container=libpod'
Signed-off-by: Matthew Heon <matthew.heon@gmail.com>
Closes: #540
Approved by: baude
Cull funcs from runtime_img.go which are no longer needed. Also, fix any remaining
spots that use the old image technique.
Signed-off-by: baude <bbaude@redhat.com>
Closes: #532
Approved by: mheon
This will allow programs to easily identify they are running in a
container
Signed-off-by: Matthew Heon <matthew.heon@gmail.com>
Closes: #533
Approved by: baude
An image name is really just a tag. When an image has multiple tags, we should be
able to "delete" the one of its tags without harm. In this case, the "delete' is
really a form of Untag (removing the tag from the image).
If an image has multiple tags and the user tries to delete by ID without force, this
should be denied because when you delete by ID there is no distinguishing it like
image tags.
Signed-off-by: baude <bbaude@redhat.com>
Closes: #528
Approved by: mheon
Migrate the podman create and commit subcommandis to leverage the images library. I also had
to migrate the cmd/ portions of run and rmi.
Signed-off-by: baude <bbaude@redhat.com>
Closes: #498
Approved by: mheon
This solves our prior problems with attach races by ensuring the
order is correct.
Also contains substantial cleanups to the attach code.
Signed-off-by: Matthew Heon <matthew.heon@gmail.com>
Closes: #482
Approved by: baude
This represents the stage3 implementation for the image library. At this point, we
are moving the image-centric functions to pkg/image including migration of args and
object-oriented references. This is a not a one-for-one migration of funcs and some
funcs will need to continue to reside in runtime_img as they are overly specific to
libpod and probably not useful to others.
Signed-off-by: baude <bbaude@redhat.com>
Closes: #484
Approved by: baude
Return and print the correct list of images by adding all specified
RepoTags to one image object, and priting them separately in
repository:repotag pairs.
Signed-off-by: Valentin Rothberg <vrothberg@suse.com>
Closes: #477
Approved by: rhatdan
Separate Init() and Start() does not make sense on the pod side,
where we may have to start containers in order to initialize
others due to dependency orders.
Also adjusts internal containers API for more code sharing.
Signed-off-by: Matthew Heon <matthew.heon@gmail.com>
Closes: #478
Approved by: rhatdan
This is the stage 1 effort for an image library that can be eventually used by buildah and
podman alike. In eventuality, the main goal of the library (package) is to:
* provide a consistent approach to resolving image names in various forms (from users).
* based on the result of the above, provide image methods that in a singular spot but separate from the runtime.
* reduce the cruft and bloat in the current podman runtime.
The goal of stage 1 is to demonstrate fast, accurate image resolution for both local and remote images resulting in
an image object as part of the return.
Signed-off-by: baude <bbaude@redhat.com>
Closes: #463
Approved by: baude
Refactors creation of bind mounts into a separate function that
can be called from elsewhere (e.g. pod start or container
restart). This function stores the mounts in the DB using the
field established last commit.
Spec generation now relies upon this field in the DB instead of
manually enumerating files to be bind mounted in.
Signed-off-by: Matthew Heon <matthew.heon@gmail.com>
Closes: #462
Approved by: baude
We don't want this in our public API - better to let us control
what gets put in container storage and where.
Signed-off-by: Matthew Heon <matthew.heon@gmail.com>
Closes: #462
Approved by: baude
The progress should not be show for import, load, and commit. It makes machine
parsing of the output much more difficult. Also, each command should output an
image ID or name for the user.
Added a --verbose flag for users that still want to see progress.
Resolves issue #450
Signed-off-by: baude <bbaude@redhat.com>
Closes: #456
Approved by: rhatdan
Notify conmon when the terminal size changes. Use the same notification
to set the correct initial size.
Closes: https://github.com/projectatomic/libpod/issues/351
Signed-off-by: Giuseppe Scrivano <gscrivan@redhat.com>
Closes: #448
Approved by: baude
The standard config has moved to /usr/share/containers/ per
discussion. An override configuration file is allowed at the
previous /etc/containers/ location. This override will be used in
place of the normal config if both are present, and exists to
override distro packaged configs without modifying the standard
config.
Signed-off-by: Matthew Heon <matthew.heon@gmail.com>
Closes: #430
Approved by: rhatdan
This allows more graceful handling of multiple paths in a config
file.
Signed-off-by: Matthew Heon <matthew.heon@gmail.com>
Closes: #430
Approved by: rhatdan
Took duplicated code and merged it into the helper function so only a single
exec was executed.
Signed-off-by: baude <bbaude@redhat.com>
Closes: #446
Approved by: mheon
Until https://github.com/containernetworking/plugins/pull/75 is merged
upstream, we are using iptables to manually allow DNS resolution in
containers that run bridged mode networking. We also remove the rule
in the networkwork tear down.
Resolves issue: #390
Signed-off-by: baude <bbaude@redhat.com>
Replace our old IP and Subnet fields in state with CNI types that
contain a lot more information. Retrieve these structs from the
CNI plugins themselves.
Signed-off-by: Matthew Heon <matthew.heon@gmail.com>
Closes: #440
Approved by: baude
Erroring can cause us to get into an state where a container
which has no exit file cannot be shown in PS, cannot be removed,
etc.
Signed-off-by: Matthew Heon <matthew.heon@gmail.com>
Closes: #438
Approved by: rhatdan
This will behave better if we need to add anything to it at a
later date - we can add fields to the struct without breaking
existing BoltDB databases.
Signed-off-by: Matthew Heon <matthew.heon@gmail.com>
Closes: #412
Approved by: baude
This allows containers to be used by `ps` and other commands
while they have ongoing exec sessions. Concurrent exec should
also work but is not tested.
Signed-off-by: Matthew Heon <matthew.heon@gmail.com>
Closes: #412
Approved by: baude
For containers without --force set, an error will be returned
For containers with --force, all pids in the container will be
stopped, first with SIGTERM and then with SIGKILL after a timeout
(this mimics the behavior of stopping a container).
Signed-off-by: Matthew Heon <matthew.heon@gmail.com>
Closes: #412
Approved by: baude
This ensures that containers with active exec sessions will not
have storage unmounted under them or network namespaces destroyed
Signed-off-by: Matthew Heon <matthew.heon@gmail.com>
Closes: #412
Approved by: baude
Exec sessions now have an ID generated and assigned to their PID
and stored in the database state. This allows us to track what
exec sessions are currently active.
Signed-off-by: Matthew Heon <matthew.heon@gmail.com>
Closes: #412
Approved by: baude
Fix an error message, and always set Privileged if the container
is also privileged.
Signed-off-by: Matthew Heon <matthew.heon@gmail.com>
Closes: #412
Approved by: baude
Currently, we will error if the DB is configured with the default
containers/storage config, and then opened by a libpod which has
explicitly set the defaults. This is due to us using an empty
config by default (to tell c/storage to use its defaults).
This patch changes our handling so that unset storage config
(using the default) and explicitly setting the defaults are both
compatible.
Signed-off-by: Matthew Heon <matthew.heon@gmail.com>
Closes: #423
Approved by: baude
add registries and insecure registries to the podman info output. this gives us
some capatibility with other container runtimes.
Resolves issue #420
Signed-off-by: baude <bbaude@redhat.com>
Closes: #422
Approved by: mheon
Add networking information to podman stats output. Also correct an issue filed
where memory constraints of the cgroup were not reflected in the stats output. And
finally, fix issue with PID count.
Resolves issue #364
Signed-off-by: baude <bbaude@redhat.com>
Closes: #417
Approved by: mheon
Also consider "/usr/lib/cni" as a potential directory for CNI plugins.
On some distributions, e.g., on openSUSE, %{_libexecdir} evaluates to
"/usr/lib".
Signed-off-by: Valentin Rothberg <vrothberg@suse.com>
Closes: #416
Approved by: mheon
Due to the way ps arguments work, it was possible to display pids
that dont below to the container in top output. We now filter pids
that dont belong to the container out of the output. This also means
the pid column must be present in the output or we throw an error.
This resolves issue #391
Signed-off-by: baude <bbaude@redhat.com>
Closes: #400
Approved by: rhatdan
When performing a podman load, if there were no repotags in the image, podman would panic. In
the case that the incoming image does have repotags, it should be imported as a none:none image
so it can still be used by the user.
Resolves issue #403
Signed-off-by: baude <bbaude@redhat.com>
Closes: #405
Approved by: baude
The packages in Lokesh's ppa that is currently used for podman puts the
conmon and runc binaries in /usr/lib/crio/bin/conmon and /usr/lib/cri-o-runc/sbin/runc
respectively.
Signed-off-by: baude <bbaude@redhat.com>
Closes: #406
Approved by: baude
Podman should not override users mounts with default mounts
for /etc/hostname, /etc/resolv.conf, and /etc/hosts.
Resolves issue #388
Signed-off-by: baude <bbaude@redhat.com>
Closes: #401
Approved by: mheon
When creating container storage by name, if that name is a tagged image then the storage
could not be found. We now use the image id which seems more reliable. Also added an
integration test to protect against regression.
Signed-off-by: baude <bbaude@redhat.com>
Closes: #393
Approved by: mheon
--image-volumes tells podman what to do with the image volumes in the image config
There are 3 options: bind, tmpfs, and ignore
bind puts the volume contents in /var/lib/containers/storage/container-id/volumes/vol-dir
and bind mounts it into the container at /vol-dir
tmpfs mounts /vol-dir as a tmps into the container
ignore doesn't mount the image volumes onto the container
Signed-off-by: umohnani8 <umohnani@redhat.com>
Closes: #377
Approved by: rhatdan
Previous code was using slow routines to collect some of the information
needed to output images. Specifically size was being calculated instead
of using the cached, already known size already available. Also, straight-
lined several of the code paths. Overall assessment is that these
improvements cut the time for images in half.
Signed-off-by: baude <bbaude@redhat.com>
Closes: #365
Approved by: mheon
Change logic for refreshing our state using runc to only poll
for conmon exit files when we first transition to the Stopped
state. After that, we should already have the exit code stored in
the database, so we don't need to look it up again.
Signed-off-by: Matthew Heon <matthew.heon@gmail.com>
Closes: #363
Approved by: TomSweeneyRedHat
Currently we unconditionally roll back transactions after error,
even if a commit has already been attempted. Commit is guaranteed
to end a transaction, though, whether by successfully committing
or by rolling back if that fails. As such, we attempt a double
rollback if a transaction fails at commit (for example, for a
constraint violation), which doesn't error but does log angry
warning messages. Ensure we don't try rolling back after commit
runs to prevent this.
Signed-off-by: Matthew Heon <matthew.heon@gmail.com>
Closes: #327
Approved by: rhatdan
Now, we don't need to use the global ID registry to iterate - we
can iterate only through containers or only through pods, without
having to iterate through both.
Signed-off-by: Matthew Heon <matthew.heon@gmail.com>
Closes: #184
Approved by: baude
This solves some dependency problems in the state, and makes
sense from a design standpoint.
Containers not in a pod can still depend on the namespaces of
containers joined to a pod, which we might also want to change in
the future.
Signed-off-by: Matthew Heon <matthew.heon@gmail.com>
Closes: #184
Approved by: baude
This allows us to JSON it and stuff it in the DB - previously,
all pod fields were private, so JSON couldn't encode them. This
allows us to keep all pod fields private by having a substruct
with public fields.
Signed-off-by: Matthew Heon <matthew.heon@gmail.com>
Closes: #184
Approved by: baude
Changing these fields caused the output of podman inspect to more
closely match docker inspect.
Signed-off-by: Daniel J Walsh <dwalsh@redhat.com>
Closes: #306
Approved by: mheon
Rename finished_amd64 to finished64.go to more accurately reflect
that it covers all 64bit arches.
Also, bumped the EPOCH for gitvalidation to speed up validations.
Signed-off-by: baude <bbaude@redhat.com>
Closes: #318
Approved by: mheon
This ensures that there is only one canonical place where
containers in a pod are stored, in the state itself.
Signed-off-by: Matthew Heon <matthew.heon@gmail.com>
Closes: #268
Approved by: rhatdan
In order to have sd_notify from systemd to work in containers
we need to pass down the NOTIFY_SOCKET environment variable to
the container.
LISTEN_FDS, tells the application inside of the container to use
socket activation and grab the FDS that are leaked into the container.
Signed-off-by: Daniel J Walsh <dwalsh@redhat.com>
Closes: #271
Approved by: umohnani8
Migrate create and commit bats tests to the ginkgo
test suite. In doing so, some structures had to be
moved to pkg/podmanstructs/podmanstructs.go so we
could do better verification of test results.
Signed-off-by: baude <bbaude@redhat.com>
Closes: #286
Approved by: rhatdan
To account for more path possibilities, we now iterate
a string array of possible paths to try and find paths
to runc and conmon.
Signed-off-by: baude <bbaude@redhat.com>
Closes: #278
Approved by: baude
Migrate ps, pull, push, and rm from bats to ginkgo.
Also, fixed a conditional issue with adding ports
when an image defines the port and the user wants
to override it.
Signed-off-by: baude <bbaude@redhat.com>
Closes: #277
Approved by: baude
Ubuntu installs runc to /usr/sbin/runc so we now account
for that. Also, added small check when creating a new
runtime that if we cannot find the runc binary, we bail
out.
Signed-off-by: baude <bbaude@redhat.com>
Closes: #276
Approved by: baude
The container's hostname should be set as an environment
variable for the container.
Signed-off-by: baude <bbaude@redhat.com>
Closes: #273
Approved by: baude
Normal Stop should not need a timeout, and should use the default
Add a function that does accept a timeout aside it
Signed-off-by: Matthew Heon <mheon@redhat.com>
Closes: #272
Approved by: rhatdan
QE found issues with formatting the go template and
the man page was lacking information.
Changed the format of the output to match latest docker.
Add shortID function that returns the truncated ID
Signed-off-by: umohnani8 <umohnani@redhat.com>
Closes: #258
Approved by: rhatdan
When trying to determine if a user-provided string that describes
an image (ID, fq name, shortname, tagged), there were some
inefficiencies where we looked up images multiple times to derive
information about local images.
Signed-off-by: baude <bbaude@redhat.com>
Matt Heon and I found that a defer statement was costing podman
run dearly. We dont think the defer function was working (nor
needed) and was timing out as well. Removing this defer statement
decreased podman runtime by 1.5s or more.
Signed-off-by: baude <bbaude@redhat.com>
Closes: #253
Approved by: baude
QE pointed out a few things missing/wrong with ps
This PR addresses those issues.
Added functionality for getting mounts and size also
Fixed a few issues with the --filter params, for
example filter with partial information.
Signed-off-by: umohnani8 <umohnani@redhat.com>
Closes: #250
Approved by: rhatdan
When an image has a port to expose, we need to expose it. User's input overrides the
image's port information.
Also, enable port information in ps so we can see which random port is assigned.
Signed-off-by: baude <bbaude@redhat.com>
Closes: #249
Approved by: rhatdan
Adds the ability to override the container's hostname. Also, uses
the first twelve characters of the container ID as the default hostname
if none is provided.
Signed-off-by: baude <bbaude@redhat.com>
Closes: #248
Approved by: baude
Set up nbetworking ports for the following use cases:
* bind the same port between host and container
* bind a specific host port to a different container port
* bind a random host port to a specific container port
Signed-off-by: baude <bbaude@redhat.com>
Closes: #214
Approved by: baude
Fix errors when containers are not running.
--all, --latest, containers can not be used at same time.
Should match the output of docker stats, 0 values replaced by "--"
Should return stats right away if container is not running.
Signed-off-by: Daniel J Walsh <dwalsh@redhat.com>
Closes: #244
Approved by: TomSweeneyRedHat
Each of these options are destructive in nature, meaning if the user
adds one of them, all current ones are removed from the produced
resolv.conf.
* dns-server allows the user to specify dns servers.
* dns-opt allows the user to specify special resolv.conf options
* dns-search allows the user to specify search domains
The add-host option is not destructive and truly just adds the host
to /etc/hosts.
Signed-off-by: baude <bbaude@redhat.com>
Closes: #231
Approved by: mheon
Weighing in at ~1700 lines, container.go is just too big. Split
it into three files: core structs and accessors (container.go),
public API (container_api.go), and internal functions
(container_internal.go).
Signed-off-by: Matthew Heon <matthew.heon@gmail.com>
If user does not specify seccomp file or seccomp file does not exist,
then use the default seccomp settings.
Still need to not hard code /etc/crio/seccomp.json, should move this to
/usr/share/seccomp/seccomp.json
Signed-off-by: Daniel J Walsh <dwalsh@redhat.com>
Closes: #233
Approved by: baude
Add new functions to update pods and add/remove containers from them
Use these new functions in place of manually modifying pods
Signed-off-by: Matthew Heon <matthew.heon@gmail.com>
Closes: #229
Approved by: rhatdan
This won't matter during batched operatins, but if the container
leaks outside of the Batch() function it will segfault if asked
to do any operation that locks unless this is applied
Signed-off-by: Matthew Heon <matthew.heon@gmail.com>
Closes: #226
Approved by: rhatdan
Also prevent containers with dependencies from being removed from
in memory states. SQLite already enforced this via FOREIGN KEY
constraints.
Signed-off-by: Matthew Heon <matthew.heon@gmail.com>
Closes: #220
Approved by: rhatdan
Remove existing code for sharing namespaces and replace with use
of this API
Signed-off-by: Matthew Heon <matthew.heon@gmail.com>
Closes: #220
Approved by: rhatdan
If we start a container and it does not error, we can assume the
container is now running. Subsequent API calls will sync for us
to see if it died, so we can just set ContainerStateRunning
instead of launching the runtime to explicitly get state.
The same logic applies to pause and unpause.
Signed-off-by: Matthew Heon <matthew.heon@gmail.com>
Closes: #223
Approved by: rhatdan
Also moves port mappings out of the SQL DB and into a file on
disk. These could get very sizable (hundred to thousands of
ports) so moving them out to a file will keep the DB small and
fast.
Finally, add a foreign key reference from container ID to
container state ID. This ensures we never get into an
inconsistent state where we have data in one table but not the
other.
Signed-off-by: Matthew Heon <matthew.heon@gmail.com>
Closes: #225
Approved by: baude
Disabling locking/syncing in a batched operation not yet implemented
Signed-off-by: Matthew Heon <matthew.heon@gmail.com>
Closes: #222
Approved by: rhatdan
It is desirable to have a --latest switch on the podman wait
command so we can wait on the latest container created to end.
Also, fixes a panic with latest where no containers are available.
Signed-off-by: baude <bbaude@redhat.com>
Closes: #201
Approved by: baude
This can now be handled by CNI plugins, so let them manage ports
instead.
Signed-off-by: Matthew Heon <matthew.heon@gmail.com>
Closes: #189
Approved by: mheon
With certain short name usages, rmi still was unable to delete
certain images. This was also reflected in several commit tests
that were temporarily disabled.
Signed-off-by: baude <bbaude@redhat.com>
Closes: #200
Approved by: rhatdan
It is desirable to have a shortcut for the most
recently created container. We can now use "**latest"
to represent the most recent container instead of its
container ID or name. For example:
Signed-off-by: baude <bbaude@redhat.com>
Closes: #179
Approved by: baude
Removing by shortname was not working. Also pruned
container storage's remove func from rmi and moved it into
an image.Remove func, which consolidates our usage of cs.
Signed-off-by: baude <bbaude@redhat.com>
Closes: #188
Approved by: baude
While pulling by shortname (fedora-minimal) worked, running a container
by the short name did not due to a logic error.
Signed-off-by: baude <bbaude@redhat.com>
Closes: #182
Approved by: rhatdan
This should help with performance when executing many operations
on a single container
Signed-off-by: Matthew Heon <matthew.heon@gmail.com>
Closes: #185
Approved by: rhatdan
We don't want libkpod overrides for conmon's path to misdirect
the already set path for conmon from libpod.
Signed-off-by: baude <bbaude@redhat.com>
Closes: #181
Approved by: baude
We should be pulling information out of the image to set the
defaults to use when setting up the container.
Signed-off-by: Daniel J Walsh <dwalsh@redhat.com>
Closes: #110
Approved by: mheon
In cases, like Ubuntu, where it uses systemd resolve
for DNS then do not copy /etc/resolv.conf but instead
the resolv.conf in the systemd resolve /run dir.
Signed-off-by: baude <bbaude@redhat.com>
Closes: #177
Approved by: rhatdan
podman run/create have the ability to set the stop timeout flag.
We need to stop it in the database.
Also Allowing negative time for stop timeout makes no sense, so switching
to timeout of uint, allows user to specify huge timeout values.
Signed-off-by: Daniel J Walsh <dwalsh@redhat.com>
Closes: #158
Approved by: TomSweeneyRedHat
podman commit allows the user to commit containers
as images with options of tagging th image, setting
a commit message, setting the auther, and making
changes to the instructions.
Signed-off-by: umohnani8 <umohnani@redhat.com>
Closes: #143
Approved by: rhatdan
Stop Signal from kpod create/run was not fully plumbed in,
This will pass the stopsignal into the container database on
create and run of containers.
Signed-off-by: Daniel J Walsh <dwalsh@redhat.com>
Closes: #156
Approved by: mheon
Also add --quiet option to kpod create/run since
this will help with writing tests.
Signed-off-by: Daniel J Walsh <dwalsh@redhat.com>
Closes: #140
Approved by: TomSweeneyRedHat
User can select from 3 manifest types: oci, v2s1, or v2s2
e.g kpod push --format v2s2 alpine dir:my-directory
Added "compress" flag to enable compression when true
Signed-off-by: umohnani8 <umohnani@redhat.com>
Closes: #126
Approved by: rhatdan
Given that we don't have a good way of cleaning up locks, these
could potential cause issues if we ever reuse a container or pod
ID
Also changes locks dir to use tmpfs, as we can't directly clean
up locks
Signed-off-by: Matthew Heon <matthew.heon@gmail.com>
Closes: #138
Approved by: rhatdan
Also includes misc other fixes - adding labels, fixing pod names
Signed-off-by: Matthew Heon <matthew.heon@gmail.com>
Closes: #138
Approved by: rhatdan
For DNS to work properly, we need to copy the host's /etc/resolv.conf
into the container during Init(). We do this by copying it into the
containers rundir and then bind mounting it into the container.
Signed-off-by: baude <bbaude@redhat.com>
Closes: #130
Approved by: baude
kpod inspect now uses the new libpod container state
and closely matches the output of docker inspect
some aspects of it are still WIP as the libpod container state
is still being worked on
Signed-off-by: umohnani8 <umohnani@redhat.com>
When loading an image, kpod load would print something like
"Trying to pull docker.io/library/alpine...", which is misleading
and makes it sound like its pulling it form the registry.
Fixed this by removing these print statements for kpod load
Signed-off-by: umohnani8 <umohnani@redhat.com>
Initial wiring of kpod exec. We wont support the following options
for exec:
* detach -- unsure of use case
* detach-keys -- not supported by runc
* interactive -- all terminals will be interactive
Not adding exec tests as we need to think about how to support a
test that requires console access but our CI tests have no console.
Signed-off-by: baude <bbaude@redhat.com>
Create an artifacts directory in the container's
static directory so store container information
coming from outside of libpod to specified files
An example is to hold data from user specified flags
in kpod run/create such as --cap-add, --ipcMode, etc...
Signed-off-by: umohnani8 <umohnani@redhat.com>
Closes: #108
Approved by: mheon
There are still two places that don't use the new function,
export and mount, but both can probably be converted to it
in the future.
Signed-off-by: Matthew Heon <matthew.heon@gmail.com>
Closes: #99
Approved by: rhatdan
Allow kpod create/run to create contianers in different network namespaces, uts namespaces and
IPC Namespaces.
This patch just handles the simple join the host, or another containers namespaces.
Lots more work needed to full integrate --net
Signed-off-by: Daniel J Walsh <dwalsh@redhat.com>
Closes: #64
Approved by: mheon
This ensures we don't open a DB with an earlier schema or a
config that differs from ours
Signed-off-by: Matthew Heon <matthew.heon@gmail.com>
Closes: #86
Approved by: rhatdan
Also migrates kpod kill and kpod stop to libpod to use the new code
Fixes force removing containers, and actually deletes containers in runc when
removing them
Start is now capable of starting even when the container is unmounted
Signed-off-by: Matthew Heon <matthew.heon@gmail.com>
Closes: #68
Approved by: rhatdan
As reported in Issue #50, we need to be able to run
or create a container based on an image ID (as well
as name).
Signed-off-by: baude <bbaude@redhat.com>
Closes: #76
Approved by: rhatdan
The new state for containers has been added
moved kpod mount and umount over to use it
Signed-off-by: Urvashi Mohnani <umohnani@redhat.com>
Closes: #57
Approved by: rhatdan
Wire this in to all state-bound container operations to ensure
syncronization of container state.
Also exposes PID of running containers via API.
Signed-off-by: Matthew Heon <matthew.heon@gmail.com>
Closes: #56
Approved by: rhatdan
The PR contains several enhancements to our CI testing.
- enable lint testing on Fedora
- add Centos Atomic as test platform
- integration tests on run on the OS natively (uncontainerized)
- builds are done in containers
- inclusion of Vagrant file for local testing
Signed-off-by: baude <bbaude@redhat.com>
Closes: #18
Approved by: mheon
Re-order the startup of a new container via run from
initialize > start > attach to initialize > attach > start.
This fixes output when running:
kpod run -i -t IMAGE command
and
kpod run IMAGE command
Signed-off-by: baude <bbaude@redhat.com>
Fixed the logic where we observed different performance
results when running an image by its fqname vs a shortname. In
the case of the latter, we resolve the name without using the
network.
Signed-off-by: baude <bbaude@redhat.com>
Closes: #37
Approved by: rhatdan
When setting the `kpod load -q` flag, output was still going to the
screen. This patch adds a check to not output to the terminal unless
there is an io.Writer specified, and then to write to the io.Writer.
Signed-off-by: Daniel J Walsh <dwalsh@redhat.com>
Closes: #23
Approved by: rhatdan
This should turn on handling of SELinux, NoNewPrivs, seccomp and Apparmor
Signed-off-by: Daniel J Walsh <dwalsh@redhat.com>
Closes: #15
Approved by: rhatdan
Matthew Heon
Nalin Dahyabhai
umohnani8
baude
W. Trevor King
Giuseppe Scrivano
Daniel J Walsh
Matthew Heon
Nathan Williams
TomSweeneyRedHat
Valentin Rothberg
Jhon Honce
Suraj Deshmukh