Add liveness check against /state #406
Merged
Conversation
This file contains hidden or 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
k8s-ci-robot
added
the
cncf-cla: yes
k8s-ci-robot
added
the
size/S
tyrannasaurusbanks
force-pushed
the
master
branch
from
ac6ca19 to
2273adc
Compare
tyrannasaurusbanks
force-pushed
the
master
branch
from
2273adc to
dc9a27d
Compare
k8s-ci-robot
added
size/M
tyrannasaurusbanks
force-pushed
the
master
branch
from
dc9a27d to
ba27b84
Compare
tyrannasaurusbanks
changed the title
|
@tyrannasaurusbanks rather than overloading |
|
I am impacted by this. Manually pulled in your changes and its holding stable atm. |
tyrannasaurusbanks
force-pushed
the
master
branch
2 times, most recently
from
414d7e7 to
be85bfa
Compare
pkg/controller/alb-controller.go
Outdated
| @@ -356,6 +360,18 @@ func (ac *albController) StatusHandler(w http.ResponseWriter, r *http.Request) { | |||
| encoder.Encode(checkResults) | |||
| } | |||
|
|
|||
| // AliveHandler only returns a empty response. It checks nothing downstream & should only used to check the controller is still running. | |||
| func (ac *albController) AliveHandler(w http.ResponseWriter, r *http.Request) { | |||
|
|
|||
There was a problem hiding this comment.
Would you prefer to see a minimal internal check added here - like maybe attain a mutex.RLock() like in /state?
Add liveness check into deployment.yaml. Add explicit responseCode to stateHandler. Also fix some compiler warnings in alb-controller and correct log statement in ec2.
tyrannasaurusbanks
force-pushed
the
master
branch
from
be85bfa to
29a4f68
Compare
bigkraig
added a commit
that referenced
this pull request
Oct 1, 2018
Add liveness check against /state
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.
What
Why
We’ve been running an older version of the controller which was working fine for us but recently we starting seeing the liveness and readiness probes start to sporadically fail, with the liveness probe causing the pod to restart. My guess is that we're being rate limited by AWS and so the API calls in /healthz are failing/retrying.
We've mitigated this by increasing the timeouts for both our liveness and readiness probes manually and increasing the period between checks.
When I went to create a PR on the project to make the probe timeout configurable, I noticed that the liveness probe has since been removed.
I'm unsure as to why, but I guess people may have run into similar issues where the liveness probe failures could cause the pod to be killed.
We feel that the liveness probe shouldn't be checking downstream dependencies (AWS APIs) through the /healthz endpoint since if these downstream dependencies fail then all that will happen is that the controller will be restarted by kube - which would leave it in no better state. I propose that it should instead just conduct a simple check to see if the controller is alive.
Notes
I chose to copy the query-param approach from /healthz onto the /state handler - shout if you'd prefer me to refactor /healthz to support liveness checking without downstream calls.
If people hate this addition, or if there's some reason I'm missing as to why there is no liveness check then let me know and I can amend the PR to simply include the timeout and logging change.
Thanks for a great project, great to see the recently activity & migration to kubernetes-sig!