Deliver test images to CAPV nodes to avoid rate limit issue #6654
Conversation
XinShuYang
commented
Sep 5, 2024
•
XinShuYang
commented
- Increase the disk size of CAPV nodes for image storage.
- Deliver test images from the image node to CAPV nodes.
|
/test-all |
|
@XinShuYang Can we move this forward? Is it helping bypass rate limiting for pull? |
Sure. I believe this workaround can resolve the current rate limit issue. Since we've logged into the Docker account on the image node, there should be no pull rate limit when pulling images on this node. We can then distribute the necessary images to any new created nodes in the CAPV cluster. |
|
/test-all |
XinShuYang
force-pushed
the
vmcratelimit
branch
from
d724f05
to
05765e1
Compare
|
/test-all |
| IP=$2 | ||
| tag=$3 | ||
|
|
||
| docker pull ${image} && docker save -o image.tar ${image} |
There was a problem hiding this comment.
After copy to capv nodes , need to clean up local tar file i.e image.tar , also can inspect the image if we want to avoid same image pull again !
There was a problem hiding this comment.
Updated the implementation to pull the same image only once.
XinShuYang
force-pushed
the
vmcratelimit
branch
from
05765e1
to
eb5ef3c
Compare
|
/test-all |
There is a limit for authenticated users, it's 200 pulls per 6 hour period. If we can get our Sponsored OSS status back, then there will be no limit. |
|
@XinShuYang all the Jenkins jobs have been stuck waiting for an executor |
* Increase the disk size of CAPV nodes for image storage. * Deliver test images from the image node to CAPV nodes. Signed-off-by: Shuyang Xin <gavinx@vmware.com>
XinShuYang
force-pushed
the
vmcratelimit
branch
from
eb5ef3c
to
c6a1eec
Compare
The Sonobuoy test is stuck because its default image pull policy is still |
|
/test-all |
