- remove 'NO TESTS NEEDED' as a valid bypass string. Henceforth
only 'NO NEW TESTS NEEDED' will work.
- add a debugging aid for #11871, in which bodhi tests time out
in nslookup.
Signed-off-by: Ed Santiago <santiago@redhat.com>
the --cgroups=split test changes the current cgroup as it creates a
sub-cgroup. This can cause a race condition in tests that are reading
the current cgroup.
Closes: https://github.com/containers/podman/issues/11191
Signed-off-by: Giuseppe Scrivano <gscrivan@redhat.com>
Instead using the OS-wide system default policy, use
the one in this repo, and adjust the expected results
(as well as making the test stricter).
Signed-off-by: Miloslav Trmač <mitr@redhat.com>
This is a cosmetic change. The help message for `podman version` is in
title case whereas all other command help messages are not in title
case. This stands out as inconsistent when looking at the output of
`podman help`.
Signed-off-by: Praveen Kumar <praveen+git@kumar.in>
Previously this test used an ad-hoc timeout mechanism to synchronize
with output of the container ID. However, depending on runtime
conditions this may not correctly correspond with complete startup
of the systemd process. Consequently this test fails under some
conditions with an error like:
`System has not been booted with systemd as init system (PID 1). Can't
operate. Failed to connect to bus: Host is down`
Fix this by using the more appropriate `WaitContainerReady()`
against output from system startup, close to finalization. In this way,
the test status command cannot run until systemd is fully operational.
Signed-off-by: Chris Evich <cevich@redhat.com>
Duplicate Address Detection slows the ipv6 setup down for 1-2 seconds.
Since slirp4netns is run it is own namespace and not directly routed
we can skip this to make the ipv6 address immediately available.
We change the default to make sure the slirp tap interface gets the
correct value assigned so DAD is disabled for it.
Also make sure to change this value back to the original after slirp4netns
is ready in case users rely on this sysctl.
Fixes#11062
Signed-off-by: Paul Holzinger <pholzing@redhat.com>
The following problems regarding `logs --tail` with the journald log
driver are fixed:
- One more line than a specified value is displayed.
- '--tail 0' displays all lines while the other log drivers displays
nothing.
- Partial lines are not considered.
- If the journald events backend is used and a container has exited,
nothing is displayed.
Integration tests that should have detected the bugs are also fixed. The
tests are executed with json-file log driver three times without this
fix.
Signed-off-by: Hironori Shiina <shiina.hironori@jp.fujitsu.com>
On Docker this is ignored, and it should be on Podman as
well. This is documented in the man page.
Fixes: https://github.com/containers/podman/issues/12002
Signed-off-by: Daniel J Walsh <dwalsh@redhat.com>
Made changes so that if the pod contains all exited containers and only infra is running, remove the pod.
resolves#11713
Signed-off-by: cdoern <cdoern@redhat.com>
If podman uses Workdir="/" or the workdir specified in the image, it
should not add it to the yaml.
If Podman find environment variables in the image, they should not
get added to the yaml.
If the container or pod do not have changes to SELinux we should not
print seLinuxOpt{}
If the container or pod do not change any dns options the yaml should
not have a dnsOption={}
If the container is not privileged it should not have privileged=false
in the yaml.
Fixes: https://github.com/containers/podman/issues/11995
Signed-off-by: Daniel J Walsh <dwalsh@redhat.com>
When run as rootless the podman network reload command tries to reload
the rootlessport ports because the childIP could have changed.
However if the containers has no ports we should skip this instead of
printing a warning.
Signed-off-by: Paul Holzinger <pholzing@redhat.com>
When we create a pod we have to parse the network mode form the config
file. This is a regression in commit d28e85741f.
Fixes#12207
Signed-off-by: Paul Holzinger <pholzing@redhat.com>
Make Podman more tolerant when parsing image volumes during container
creation and further fix an infinite loop when checking them.
Consider `VOLUME ['/etc/foo', '/etc/bar']` in a Containerfile. While
it looks correct to the human eye, the single quotes are wrong and yield
the two volumes to be `[/etc/foo,` and `/etc/bar]` in Podman and Docker.
When running the container, it'll create a directory `bar]` in `/etc`
and a directory `[` in `/` with two subdirectories `etc/foo,`. This
behavior is surprising to me but how Docker behaves. We may improve on
that in the future. Note that the correct way to syntax for volumes in
a Containerfile is `VOLUME /A /B /C` or `VOLUME ["/A", "/B", "/C"]`;
single quotes are not supported.
This change restores this behavior without breaking container creation
or ending up in an infinite loop.
BZ: https://bugzilla.redhat.com/show_bug.cgi?id=2014149
Signed-off-by: Valentin Rothberg <rothberg@redhat.com>
Following commit makes sure when buildah tag is invoked on a manifest
list, it tags the same manifest list instead of resolving to an image and
tagging it.
Backporting https://github.com/containers/podman/pull/12057
Signed-off-by: Aditya Rajan <arajan@redhat.com>
Multiarch folks are seeing flakes in this test. I can't reproduce
them, but I did notice that the test isn't doing the best possible
job of reporting failures nor of confirming what it purports to test.
Major fix here is to check the exit status of each curl: if we
see the flake again, that will help us track down the failure.
Other fixes are just refactoring, cleanup, and disambiguation
(using the random service name consistently)
Signed-off-by: Ed Santiago <santiago@redhat.com>
Moving to Fedora 35 showed test failures (time outs) in the test
"podman checkpoint and restore container with different port mappings"
The test starts a container and maps the internal port 6379 to the local
port 1234 ('-p 1234:6379') and then tries to connect to localhost:1234
On Fedora 35 this failed and blocked the test because the container was
not yet ready. The test was trying to connect to localhost:1234 but
nothing was running there. So the error was not checkpointing related.
Before trying to connect to the container the test is now waiting for
the container to be ready.
Another problem with this test and running ginkgo in parallel was that
it was possible that the port was already in use. Now for each run a
random port is selected to decrease the chance of collisions.
Signed-off-by: Adrian Reber <areber@redhat.com>
When the targetPort is not defined, it is supposed to
be set to the port value according to the k8s docs.
Add tests for targetPort.
Update tests to be able to check the Service yaml that
is generated.
Signed-off-by: Urvashi Mohnani <umohnani@redhat.com>
It was observed during initial F35 testing, this test can cause Ginkgo
to "hang" by attempting to connect before the redis is up/listening.
Fix this by confirming the ready-state before attempting to connect.
Also, force IPv4 and timeout on any connection fault - to allow other
tests to run.
Thanks to Adrian Reber for help on this and related fixes.
Signed-off-by: Chris Evich <cevich@redhat.com>
As the default protocol in k8s is TCP, don't add it
to the generate yaml when using protocol.
Add UDP to the protocol of the generated yaml when udp
is being used.
Add tests for this as well.
Signed-off-by: Urvashi Mohnani <umohnani@redhat.com>
If no entrypoint or command is set in the podman create
command, and the image command or entrypoint is being
used as the default, then do not add the image command or
entrypoint to the generated kube yaml.
Kubernetes knows to default to the image command and/or
entrypoint settings when not defined in the kube yaml.
Add and modify tests for this case.
Signed-off-by: Urvashi Mohnani <umohnani@redhat.com>
Removed the inclusion of RunAsUser or RunAsGroup unless a container is run with the --user flag. When building from an image
the user will be pulled from there anyway
resolves#11914
Signed-off-by: cdoern <cdoern@redhat.com>
There is a race where `conn.Close()` was called before `conn.CloseWrite()`.
In this case `CloseWrite` will fail and an useless error is printed. To
fix this we move the the `CloseWrite()` call to the same goroutine to
remove the race. This ensures that `CloseWrite()` is called before
`Close()` and never afterwards.
Also fixed podman-remote run where the STDIN was never was closed.
This is causing flakes in CI testing.
[NO TESTS NEEDED]
Fixes#11856
Signed-off-by: Paul Holzinger <pholzing@redhat.com>
<MH: Fixed cherry-pick conflicts>
Signed-off-by: Matthew Heon <matthew.heon@pm.me>
systemd sometimes spits out lines in the wrong order. Deal with it.
This fixes an infrequent flake that I haven't filed because I
didn't understand it well enough. (Hence, this reduces BUGS
but does not reduce BUG COUNT. Sorry!)
Signed-off-by: Ed Santiago <santiago@redhat.com>
...in volume test. Looks like Bodhi gating tests run from a
nonwritable directory. I feel really stupid for not realizing
this when I first tried to fix this bug two weeks ago.
Signed-off-by: Ed Santiago <santiago@redhat.com>
When using play kube and generate kube, we need to support if bind
mounts have selinux options. As kubernetes does not support selinux in
this way, we tuck the selinux values into a pod annotation for
generation of the kube yaml. Then on play, we check annotations to see
if a value for the mount exists and apply it.
Fixes BZ #1984081
Signed-off-by: Brent Baude <bbaude@redhat.com>
As we were not updating the pod ID bucket, removing a pod with
containers still in it (including the infra container, which will
always suffer from this) will not properly update the name
registry to remove the name of any renamed containers. This
patch ensures that does not happen - all containers will be fully
removed, even if renamed.
Fixes#11750
Signed-off-by: Matthew Heon <matthew.heon@pm.me>
- logs: remove unnecessary sleeps. This saves ~25s.
Unfortunately, journald seems to have some sort of lag,
so we need to keep retrying until we get the 'after' string.
- ps: add placeholder test for once buildah 3544 is fixed
- cp: bulk-kill containers when finished, instead of one by one.
This is a big change and only saves about 8s per run, but hey.
- mount,pause,healthcheck: 'podman stop -t 0' before rm'ing containers.
Easy 50s.
Have I mentioned, lately, that 'podman rm -f' needs a '-t 0' flag?
- play: same, and also 'podman pod stop'. Seems to shave ~20s.
- socket-activation: UGH! Buggy and useless tests! They were
running "sleep 90" containers for no reason whatsoever. I
assume the intention was to run them with "-d", so that's
what I've done here. Also fixed some language. 180 seconds!
(Unrelated: cleanup in 070-build, use $IMAGE, not alpine)
Signed-off-by: Ed Santiago <santiago@redhat.com>
Following commit ensures we silently return container id on `stop` if
container was never created in OCI runtime.
Following behaviour ensures that we are in parity with docker.
Signed-off-by: Aditya Rajan <arajan@redhat.com>
The shell completion should only suggest arguments that work. Using a
image without tag does not work in many cases. Having both the version
with and without tag also forces users to press one key more because
tab completion will always stop at the colon.
Fixes#11673
Signed-off-by: Paul Holzinger <pholzing@redhat.com>
If the command came from the underlying image, then we should
not include it in the generate yaml file.
Fixes: https://github.com/containers/podman/issues/11672
Signed-off-by: Daniel J Walsh <dwalsh@redhat.com>
When performing an image build with play kube, we need to set the
context directory so things like file copies have the correct input
path.
Signed-off-by: Brent Baude <bbaude@redhat.com>
Currently we add the default PATH, TERM and container from Podman
to every kubernetes.yaml file. These values should not be recorded
in the yaml files.
Signed-off-by: Daniel J Walsh <dwalsh@redhat.com>
<MH: Fixed cherry-pick conflicts>
Signed-off-by: Matthew Heon <matthew.heon@pm.me>
The health check result is stored in the container state. Since the
state can change or might not even be set we have to retrive the current
state before we try to read the health check result.
Fixes#11687
Signed-off-by: Paul Holzinger <pholzing@redhat.com>
(Sorry, couldn't resist).
CI flakes have been coming down - thank you to everyone who has
been making them a priority.
This leaves a noisy subset that I've just been ignoring for months:
Running: podman ... -p 8080:something
...cannot listen on the TCP port: listen tcp4 :8080: bind: address already in use
Sometimes these are one-time errors resolved on 2nd try; sometimes
they fail three times, forcing CI user to hit Rerun. In all cases
they make noise in my flake logs, which costs me time.
My assumption is that this has to do with ginkgo running random
tests in parallel. Since many e2e tests simplemindedly use 8080,
collisions are inevitable.
Solution: simplemindedly replace 8080 with other (also arbitrarily
picked) numbers. This is imperfect -- it requires human developers
to pick a number NNNN and 'grep NNNN test/e2e/*' before adding
new tests, which I am 100% confident ain't gonna happen -- but
it's better than what we have now.
Side note: I considered writing and using a RandomAvailablePort()
helper, but that would still be racy. Plus, it would be a pain
to interpolate strings into so many places. Finally, with this
hand-tooled approach, if/when we _do_ get conflicts on port NNNN,
it should be very easy to grep for NNNN, find the offending tests
that reuse that port, and fix one of them.
Signed-off-by: Ed Santiago <santiago@redhat.com>
psgo added support for listing supplementary groups via
two new descriptors:
* `groups` for supplementary groups inside the container
* `hgroups` for the counterpart on the host
Signed-off-by: Valentin Rothberg <rothberg@redhat.com>
- Added tests to help ensure there is no future regressions
- Added WaitWithTimeout(int) rather than calling
WaitWithDefaultTimeout() multiple times
- Exposed DefaultWaitTimeout to allow test to use a multiplier
Fixes#2221
Signed-off-by: Jhon Honce <jhonce@redhat.com>
Docker adds the `sha256:` prefix to the image ID, so our compat endpoint
has to do this as well.
Fixes#11623
Signed-off-by: Paul Holzinger <pholzing@redhat.com>
Fix a bug when remotely untagging an image via tag@digest.
The digest has been lost in the remote client and hence led
to a wrong behaviour on the server.
Fixes: #11557
Signed-off-by: Valentin Rothberg <rothberg@redhat.com>
* 070-build:
- remove workaround for #9567, which is closed.
- add many more cases to the ignorefile test,
to test complicated special cases of Buildah PR 3486.
* 160-volumes:
- remove a skip_if_remote, volumes now work on remote
- use a random name for tarball, and clean up when
done using it. This fixes a gating-test failure
(test runs as root, then rootless, and rootless
can't clobber root's file).
Signed-off-by: Ed Santiago <santiago@redhat.com>
The check for net=none was wrong. It just assumed when we do not create
the netns but have one set that we use the none mode. This however also
applies to a container which joins the pod netns.
To correctly check for the none mode use `config.NetMode.IsNone()`.
Fixes#11596
Signed-off-by: Paul Holzinger <pholzing@redhat.com>
When no name is given for podman container runlabel it will default to
the image base name. However this can contain a tag. Since podman does
not accept container names with a colon the run command will fail if it
contains something like `podman run --name NAME ...`.
Fixes https://bugzilla.redhat.com/show_bug.cgi?id=2004263
Signed-off-by: Paul Holzinger <pholzing@redhat.com>
