Move hack/travis/install-etcd.sh to hack/, and update etcd version to 2.0.12 #15336
Conversation
ixdy
force-pushed
the
install-etcd
branch
2 times, most recently
from
b45ab64 to
42838fd
Compare
| @@ -35,11 +35,12 @@ RUN rm -rf /var/lib/apt/lists/ | |||
| RUN apt-get -o Acquire::Check-Valid-Until=false update && apt-get install -y rsync | |||
|
|
|||
| # Download and symlink etcd. We need this for our integration tests. | |||
| ENV ETCD_VERSION v2.0.12 | |||
There was a problem hiding this comment.
It's important to keep this in sync with what we install in our default cluster builds. How are the two kept in sync?
k8s-github-robot
added
the
size/S
|
Labelling this PR as size/S |
|
|
||
| ETCD_VERSION=${ETCD_VERSION:-v2.0.0} | ||
| ETCD_VERSION=${ETCD_VERSION:-v2.0.12} |
There was a problem hiding this comment.
Another instance of this literal. I know it's going to be a bit messy to implement, but can we not specify it canonically somewhere?
|
GCE e2e test build/test passed for commit 76e1f284fdae3c7ea5200ccb8ba65fdd0943bfa9. |
|
Some nits, which I'm happy to defer to a later PR if necessary. Other than those, LGTM. |
|
GCE e2e test build/test passed for commit b45ab643ed5f686bb65ba508eabaacaee6e5c859. |
|
GCE e2e test build/test passed for commit 42838fdf8c043cd554d7b4a6f2470743d2c8e8e1. |
|
As you've noted, keeping all of these versions in sync right now is a complete mess. Just looking for the "current" version (2.0.12), this is explicitly referenced in many files: $ grep "2\.0\.12" -l -r *
build/build-image/Dockerfile
cluster/mesos/docker/docker-compose.yml
cluster/ubuntu/build.sh
cluster/images/etcd/Makefile
cluster/saltbase/salt/etcd/etcd.manifest
cluster/centos/config-build.sh
docs/getting-started-guides/scratch.md
docs/getting-started-guides/docker.md
docs/getting-started-guides/mesos.md
docs/getting-started-guides/ubuntu.md
docs/getting-started-guides/docker-multinode/master.md
docs/getting-started-guides/docker-multinode/master.sh
docs/devel/development.md
docs/admin/etcd.md
hack/install-etcd.sh
test/kubemark/start-kubemark-master.shIt's unclear how many other scripts are specifying a different version. The source of truth is in cluster/saltbase/salt/etcd/etcd.manifest: but besides parsing that file (and then downloading and extracting the binaries from that docker image) I'm not sure how we can easily combine all of these references. It'd be a good idea to fix this, but it seems like a harder problem. |
|
GCE e2e test build/test passed for commit 840a9158c2d9077e734645c7cc9f822b989772c7. |
|
GCE e2e test build/test passed for commit eeb745dc1e04dfb39d97dffe38cf8e8c01d44c55. |
|
FWIW, there's another whole install-etcd.sh here: https://github.com/kubernetes/kubernetes/blob/master/cluster/mesos/docker/test/bin/install-etcd.sh Ideally it'd use |
|
@karlkfi yeah, I saw that one as well at some point, though it missed my recent grep. It specifies etcd version 2.0.11 - should I update it to 2.0.12 as well? |
|
GCE e2e test build/test passed for commit e5842c5dea98cd8b65cca0e8c0a177d7b15252a4. |
|
Sure, go ahead, if you're changing all the rest. |
|
Also updated |
ixdy
changed the title
|
GCE e2e test build/test passed for commit 1c5f252. |
ghost
added
the
lgtm
|
LGTM |
Move hack/travis/install-etcd.sh to hack/, and update etcd version to 2.0.12
Move hack/travis/install-etcd.sh to hack/, and update etcd version to 2.0.12
First part is a cleanup from #13951 (comment), since both Jenkins + Shippable are using this script (and not just Travis).
Second part is catching up our local testing to match what we really use in clusters as of June (ec1e308).
@kubernetes/goog-testing