Remove public IPs from masters #1045
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
openshift-ci-robot
added
do-not-merge/work-in-progress
|
Don't we still need these until we get a better fix than openshift/origin#21700 for CI? |
|
I'm hacking on a simple tool for dev and CI (but not for customers) to get an ssh pod running which can act as an easy to deploy bastion. Maybe I'll be able to show that PoC tomorrow. I think that should be enough to merge this... |
|
Linking the bastion config support so I can find it later: openshift/release#2469 ;). |
|
/cc @blrm |
eparis
force-pushed
the
master-private
branch
2 times, most recently
from
f107702 to
c00463a
Compare
openshift-ci-robot
added
the
needs-rebase
eparis
force-pushed
the
master-private
branch
from
c00463a to
8052f24
Compare
openshift-ci-robot
removed
the
needs-rebase
|
@wking on a slightly different topic, shouldn't the installer deploy a bastion too for operators to be able to ssh into masters/ workers ? While is all nice the work has been already done for CI, for prod deployment we leave the operator in the dark ? ... my 0.02$ though |
|
I'm in favor of pushing people away from jumping to SSH as their go-to. Providing a bastion by default goes against that goal. Creating a bastion host should be trivial for administrators and if it isn't, there are plenty of strategies that leverage SSH pods (like @eparis mentioned). |
wking
added
platform/aws
and removed
do-not-merge/work-in-progress
eparis
force-pushed
the
master-private
branch
from
8052f24 to
45d32fe
Compare
openshift-ci-robot
added
the
size/M
|
https://github.com/eparis/ssh-bastion provides an easy to deploy bastion pod. |
eparis
changed the title
|
now to watch what is still using ssh in CI. hopefully most things have moved to |
|
Holding while I wait for a little more buy-in. |
openshift-ci-robot
added
the
do-not-merge/hold
I'm sold :). /hold cancel /retest |
openshift-ci-robot
removed
the
do-not-merge/hold
|
/hold @wking not from this team specifically ;) I'll remove the hold once I'm satisfied with the responses I get (offline). |
openshift-ci-robot
added
the
do-not-merge/hold
Does this exist yet? Being able to have an |
FWIW I sometimes |
|
I understand why this change is a good idea. I think some help in the description breaking down the debugging scenarios would help everyone.
|
If we get far enough for bootstrap teardown, we were at least beyond this point at some point. So use kubelet pods on a master, or, lacking that, fall back to the bootstrap node. |
This was referenced Feb 8, 2019
|
Now that openshift/origin#21997 has merged, I think cases 1, 2, and 4 are reasonably covered. /hold cancel |
openshift-ci-robot
removed
the
do-not-merge/hold
|
/retest |
/retest |
|
e2e-aws had a number of flakes and then: /retest |
/retest |
wking
added a commit
to wking/openshift-installer
that referenced
this pull request
Feb 27, 2019
In 6c10827 (Removing unused/deprecated security groups and ports, 2019-02-23, openshift#1306), we restricted master SSH access to the cluster, catching up with 6add0ab (data/aws: move the masters to the private subnets, 2019-01-10, openshift#1045). But the bootstrap node is a useful SSH bastion for debugging hung installs (until we get far enough along to tear down the bootstrap resources). This commit restores global SSH access to the bootstrap node, now that it is no longer provided by the master security group.
This was referenced Feb 27, 2019
wking
added a commit
to wking/openshift-release
that referenced
this pull request
Apr 24, 2019
These are from b7cc916 (Set KUBE_SSH_USER for new installer for AWS tests, openshift#2274) and 43dde9e (Set KUBE_SSH_BASTION and KUBE_SSH_KEY_PATH in installer tests, 2018-12-23, openshift#2469). But moving forward, reliable SSH access direct to nodes will be hard, with things like openshift/installer@6add0ab447 (Remove public IPs from masters, 2019-01-10, openshift/installer#1045) making a SSH bastion a requirement for that sort of thing (at least on AWS). Going forward, ideally e2e tests can be ported to use privileged pods within the cluster to check what they need to check. But however that works out, stop carrying local dead code that is not affecting test results. We can always drag it back out of version control later if it turns out we actually want to go down the KUBE_SSH_* route.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This renames our subnets from master/worker to be public/private. It then moves the masters to the 'private' subnet. It also removes the public IP from all of the masters.
It leaves the bootstrap node in the public subnet and leaves it with a public ip.
This better follows AWS best security practices by not exposing our machines on the open internet. It does mean that a customer will need to provide their own bastion server for access.
Fixes: #747