Use different containerd sock address in tests #9415
Conversation
|
The address is configurable via the flag, can't we just pass the flag when we run our tests? (or locally when a user is running the tests they can provide the flag) |
Hmm isn't the -address flag for the tests mostly to know which sock address to connect to? It would be nice to have parity in how the test daemon is started when the '--no-daemon' flag is not specified. That is start on a different non-default sock address for the tests. like npipe:\containerd-containerd-test |
Also, tests like the following try to get the ttrpc address from the "address" set in TestMain() which is a unique test only sock address in 1.6 which makes more sense: |
|
If the test address is now same as the default address for the platform and if this is already open, tests would fail while trying to open same pipe |
kiashok
force-pushed
the
fixIntegrationClientAddr
branch
2 times, most recently
from
40054df
to
18a48d8
Compare
It may be (I'm reviewing on mobile so didn't do great due diligence 😅). The description made it seem like thats the addr it'd listen on if they were launching the daemon |
|
/retest |
Signed-off-by: Kirtana Ashok <kiashok@microsoft.com>
kiashok
force-pushed
the
fixIntegrationClientAddr
branch
from
18a48d8
to
c21bc30
Compare
|
PR needs rebase. Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the kubernetes/test-infra repository. |
If the containerd/containerd/defaults.DefaultAddress is used , it causes failures if containerd is already running on the default address and the tests are run without the 'no-daemon' flag.
It would be good to use a different containerd sock address for integration tests