InfraContainer should go through the same creation process as regular containers. This change was from the cmd level
down, involving new container CLI opts and specgen creating functions. What now happens is that both container and pod
cli options are populated in cmd and used to create a podSpecgen and a containerSpecgen. The process then goes as follows
FillOutSpecGen (infra) -> MapSpec (podOpts -> infraOpts) -> PodCreate -> MakePod -> createPodOptions -> NewPod -> CompleteSpec (infra) -> MakeContainer -> NewContainer -> newContainer -> AddInfra (to pod state)
Signed-off-by: cdoern <cdoern@redhat.com>
* Add response.Body.Close() where needed to release HTTP
connections to API server.
* Add tests to ensure no general leaks occur. 100% coverage would be
required to ensure no leaks on any call.
* Update code comments to be godoc correct
Signed-off-by: Jhon Honce <jhonce@redhat.com>
add the ability for play kube to tear down based on the yaml used to
play it. it is indicated by --down in the play kube command. volumes
are NOT deleted during the teardown. pods and their containers are
stopped and removed.
Signed-off-by: Brent Baude <bbaude@redhat.com>
Commit 9ac5267 changed the type of the generated systemd units from
`forking` to `notify`. It further stopped using `--cidfile` and instead
intended systemd to take care of stopping the container, which turned
out to be a bad idea.
Systemd will send the stop/kill signals to conmon which in turn may exit
non-zero, depending on the signal, and ultimately breaking container
cleanup.
Hence, we need to use --cidfile again and let podman stop and remove the
container to make sure that everything's in order.
Fixes: #11304
Signed-off-by: Valentin Rothberg <rothberg@redhat.com>
Podman inspect has to show exposed ports to match docker. This requires
storing the exposed ports in the container config.
A exposed port is shown as `"80/tcp": null` while a forwarded port is
shown as `"80/tcp": [{"HostIp": "", "HostPort": "8080" }]`.
Also make sure to add the exposed ports to the new image when the
container is commited.
Fixes#10777
Signed-off-by: Paul Holzinger <pholzing@redhat.com>
This reverts commit 70801b3d71.
It turns out that letting systemd handle stopping the container is not
working as I thought it will. Conmon is receiving the stop/kill signals
and may exit non-zero, which in turn lets the systemd service transition
into the `failed` state.
We need to get back to letting Podman stop the containers and do a
partial revert of commit 9ac5267 which removed using --cidfile.
Happening in a following commit.
Signed-off-by: Valentin Rothberg <rothberg@redhat.com>
Implement a new network interface to abstract CNI from libpod. The
interface is implemented for the CNI backend but in the future we can
add more backends.
The code is structured in three new packages:
- `libpod/network/types`: contains the interface definition
and the necessary types for it.
- `libpod/network/cni` contains the interface implementation for the CNI
backend.
- `libpod/network/util` a set of utility functions related to
networking.
The CNI package uses ginkgo style unit tests. To test Setup/Teardown the
test must be run as root. Each test will run in their own namespace to
make the test independent from the host environment.
New features with the CNI backend:
- The default network will be created in memory if it does not exists on
disk.
- It can set more than one static IP per container network.
- Networks are loaded once from disk and only if this interface is
used, e.g. for commands such as `podman info` networks are not loaded.
This reduces unnecessary disk IO.
This commit only adds the interface it is not wired into libpod. This
requires a lot of breaking changes which will be done in a followup
commit.
Once this is integrated into libpod the current network code under
`libpod/network` should be removed. Also the dependency on OCICNI
should be dropped.
Signed-off-by: Paul Holzinger <pholzing@redhat.com>
Commit 9ac5267598 changed the type of the generated systemd units from
forking to notify. Parts of these changes was also removing the need to
pass any information via the file system (e.g., PIDFILE, container ID).
That in turn implies that systemd takes care of stopping the container.
By default, systemd first sends a SIGTERM and after a certain timeout,
it'll send a SIGKILL. That's pretty much what Podman is doing, unless
the container was created with a custom stop signal which is the case
when the --stop-signal flag was used or systemd is mounted.
Account for that by using systemd's KillSignal option which allows for
changing SIGTERM to another signal. Also make sure that we're using the
correct timeout for units generated with --new.
Fixes: #11304
Signed-off-by: Valentin Rothberg <rothberg@redhat.com>
In libpod/logs.LogLine.Write(), don't write a newline to stdout/stderr
when the log message is only part of a line.
In libpod.ConmonOCIRuntime.HTTPAttach(), don't send a newline over the
HTTP connection when the log message is only part of a line.
In pkg/api/handlers/compat.LogsFromContainer(), don't send a newline
over the HTTP connection when the log message is only part of a line,
and don't make doing so conditional on whether or not the client used
the docker or podman endpoint.
In pkg/domain/infra/tunnel.ContainerEngine.ContainerLogs(), don't add
our own newline to log messages, since they already come through from
the server when they need to.
Signed-off-by: Nalin Dahyabhai <nalin@redhat.com>
we were adding a negative duration in podman events, causing inputs like
-5s to be correct and 5s to be incorrect.
fixes#11158
Signed-off-by: cdoern <cdoern@redhat.com>
Adds support for transferring data between systems and backing up systems.
Use cases: recover from disasters or move data between machines.
Signed-off-by: flouthoc <flouthoc.git@gmail.com>
Add a new flag `--memory-profile=$path` which creates a memory profile.
The generated profile can later be analyzed via `go tool pprof`.
[NO TESTS NEEDED] since it's a hidden flag, devs-only.
Signed-off-by: Valentin Rothberg <rothberg@redhat.com>
For docker compat include information about available volume, log and
network drivers which should be listed under the plugins key.
Fixes#11265
Signed-off-by: Paul Holzinger <pholzing@redhat.com>
When the rootlessport process is started the stdout/stderr are attached
to the podman process. However once everything is setup podman exits and
when the rootlessport process tries to write to stdout it will fail with
SIGPIPE. The code handles this signal and puts /dev/null to stdout and
stderr but this is not robust. I do not understand the exact cause but
sometimes the process is still killed by SIGPIPE. Either go lost the
signal or the process got already killed before the goroutine could
handle it.
Instead of handling SIGPIPE just set /dev/null to stdout and stderr
before podman exits. With this there should be no race and no way to
run into SIGPIPE errors.
[NO TESTS NEEDED]
Fixes#11248
Signed-off-by: Paul Holzinger <pholzing@redhat.com>
When setting path names in the build context archive, convert path names
to use forward slashes, as is normal for those archives, so that
directory hierarchies archived on Windows hosts extract correctly
everywhere.
Not really sure how to run the remote client in CI on a system that uses
`\` as a path separator, which is where this error crops up, so
[NO TESTS NEEDED]
Signed-off-by: Nalin Dahyabhai <nalin@redhat.com>
When playing a kube YAML file, it can be desirable to be able to build
an image on the fly. This is good for development of an image and YAML
files and somewhat mocks what compose does.
Signed-off-by: Brent Baude <bbaude@redhat.com>
Dealing with os.Signal channels seems more like an art than science
since signals may get lost. os.Notify doesn't block on an unbuffered
channel, so users are expected to know what they're doing or hope for
the best.
In the recent past, I've seen a number of flakes and BZs on non-amd64
architectures where I was under the impression that signals may got
lost, for instance, during stop and exec.
[NO TESTS NEEDED] since this is art.
Signed-off-by: Valentin Rothberg <rothberg@redhat.com>
after the init containers pr merged, it was suggested to use `once`
instead of `oneshot` containers as it is more aligned with other
terminiology used similarily.
[NO TESTS NEEDED]
Signed-off-by: Brent Baude <bbaude@redhat.com>
We have reverted the previous patches to look for the gvproxy binary in
/usr/lib/podman and have again decided to use /usr/libexec/podman
[NO TESTS NEEDED]
Signed-off-by: Brent Baude <bbaude@redhat.com>
cdoern
OpenShift Merge Robot
Jhon Honce
Brent Baude
Ashley Cui
Valentin Rothberg
Paul Holzinger
Nalin Dahyabhai
flouthoc
Guillaume Rose
openshift-ci[bot]
xatier