-
Notifications
You must be signed in to change notification settings - Fork 40k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Validate Docker 1.10 #19720
Comments
@Random-Liu Can you pick up this work along with Huawei friends? |
@dchen1107 Sure. We could also talk about this on Tuesday meeting, :) |
@Random-Liu It will be a good start to try creating a cluster with docker 1.10 and leave it over the weekend if you have time today. |
FYI: Instructions to update salt to use the new docker rc can be found here |
FYI: rc2 has been released. Linux 64bit binary: https://test.docker.com/builds/Linux/x86_64/docker-1.10.0-rc2 https://github.com/docker/docker/issues?q=milestone:1.10.0+label:status/6-cherry-picked+is:closed Any updates @Random-Liu ? |
@vishh Thank you very much~ Now the main problem is that the option "-d" is completed deprecated in docker 1.10, while our ContainerVM is still using it. @dchen1107 is helping me solve this problem, after that I'll start a cluster and test it ASAP. :) |
@Random-Liu, you can also make a one time change and run any non-reboot tests :-) |
@yujuhong Thanks, let me try it. |
With the help of @dchen1107, I finally start the cluster successfully. I'll send a PR to add more instruction in the comments of |
I don't know whether the test suite I run is correct... Docker version:
My cluster is sync up to: 3db1a6c the following result:
|
I run the e2e test a second time, and got the same result:
|
Status update: |
Are we using the official debian package for rc2? If yes, can you file an issue against docker for this? |
Looks like docker 1.10-rc2 adds a dependency to libsystemd-journal0 (>= 201) which is not required by docker 1.10-rc1 |
Yeah, I'm using the official debian package for rc2 from http://apt.dockerproject.org/repo/pool/testing/d/docker-engine/ |
FYI, I found a relative issue for docker 1.9.1 moby/moby#19230. |
No the problem is from our recent debian backport images on which we build containervm-image: The following packages have unmet dependencies: cc/ @zmerlynn |
I'm confused. container-vm has never (yet) used the official |
Ah, yeah, it looks like |
@zmerlynn It is broken for our recent ones: 1) Debian 3.16.7-ckt11-1+deb8u6~bpo70+1 (2015-11-11) But works for the old container vm image: Debian 3.16.7-ckt11-1+deb8u5 (2015-10-09) |
|
I repro'd on a pretty old CVM (
? |
@Random-Liu showed me that he has docker 1.10-rc2 works on container vm image: Debian 3.16.7-ckt11-1+deb8u5 (2015-10-09) earlier |
We're running into performance issues w/ density https://bugzilla.redhat.com/show_bug.cgi?id=1340518 /cc @rrati @jeremyeder |
"operation timeout: context deadline exceeded" would only happen if a request takes over > 2 min. |
@timothysc @yujuhong We only run benchmark for docker 1.9 before, because we finally decided to go with 1.9 at that time. #19720 (comment) |
@Random-Liu my bad. I thought we did that. |
@yujuhong The VM hung with 100% cpu usage during benchmark running...I can't access the VM now. May have to create a new environment now. |
Slightly off-topic, but do you all do testing against Docker that is packaged and maintained by certain distros, e.g. Fedora's Docker or just the official packages? From my own testing their fork has also got its own set of issues not found in the official Docker packages, but it's currently being used with |
@bacongobbler - For RHEL based platforms (RHEL, Centos, Fedora), we at Red Hat test with the package provided by our distribution, and any upstream testing feedback we provide to the Kubernetes project is against that distribution as well. This includes the recommendation that we will make for our platforms with Kubernetes 1.3 The version of docker we package in RHEL-based platforms has an explicit list of the patches we carry.
Many of those patches are fixes we backport into docker 1.10 rather than force users to wait for docker 1.11. For example, a vanilla docker 1.10 may have the following issue: But we carry the fix in our distribution rather than force users to wait/use docker 1.11: The vagrant environment on each node should currently be using the docker 1.10 from here:
The Vagrant VM represents a setup for how Kubernetes should run on that particular distribution. This means its an ideal environment to ensure we conform to the systemd node spec as it evolves, that we work properly with the docker systemd cgroup driver, and that selinux functions as expected for users that require it. If you want to change the version of docker installed in the vagrant environment, I would accept a PR that made it an option in |
Our density results again 1.10 are showing a lag when compared against upstream 1.11 results. But it don't think it's an apples to apples comparison. @gmarek what is the size of the nodes on the 100 node test cluster. /cc @kubernetes/sig-scalability |
@timothysc If you were asking about the gce-scalibility suite, below is the information from the build log.
|
@timothysc - as @yujuhong we're using n1-standard-2 machines. In bigger clusters we generally use n1-standard-1's. |
v1.9 results copied from this comment Introduction Benchmark Environment
1. Benchmark list/inspect with varying number of containersUsing 2. Benchmark list/inspect with varying operation intervalsUsing
3. Benchmark list/inspect with varying number of goroutinesUsing |
I run the benchmark, and didn't see the issue #19720 (comment). Since I've already run the benchmark, just post the data here, for fun. :) |
Thanks @Random-Liu! @timothysc, it seems like the issue may be specific to your test environment. I reassigned the issue to you for the followup. |
The BZs are up to date. We've closed 1 as an installer issue, the 2nd one, or the wedge... Is being looked into. |
@Random-Liu awesome research, thank you. |
@runcom could we cross-link to the 1.10 errata once it's done, for folks who may be tracking 1.10 and kube 1.3. |
@timothysc probably @lsm5 may know how to link to errata for docker 1.10.3 issues |
time to close this one? |
I believe so, yes. |
@timothysc, do we need any documentation for the v1.3 release or is that covered in other issues? |
Let's use #25893 track documentation. |
We have just pushed 1.10.0-rc1 to test.docker.com you can download with the following:
Ubuntu/Debian: curl -fsSL https://test.docker.com/ | sh
Linux 64bit tgz: https://test.docker.com/builds/Linux/x86_64/docker-1.10.0-rc1.tgz
IMPORTANT:
Docker 1.10 uses a new content-addressable storage for images and layers.
A migration is performed the first time docker is run, and can take a significant
amount of time depending on the number of images and containers present.
Refer to this page on the wiki for more information:
https://github.com/docker/docker/wiki/Engine-v1.10.0-content-addressability-migration
We also released a cool migration utility that enables you to perform the migration
before updating to reduce downtime. Engine 1.10 migrator can be found on Docker Hub;
https://hub.docker.com/r/docker/v1.10-migrator/
@kubernetes/sig-node
The text was updated successfully, but these errors were encountered: