Working Kube Cancel. #3983
Merged
Working Kube Cancel. #3983
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
github-actions
bot
added
area/platform
… comments to explain why.
davinchia
commented
Jun 9, 2021
| /** | ||
| * Intended to gracefully clean up after a completed Kube Pod. This should only be called if the | ||
| * process is successful. | ||
| */ | ||
| @Override |
There was a problem hiding this comment.
@jrhizor did I correctly interpret the difference between the two waitFor functions? I double checked the WorkerUtils class to confirm this is how they are used today.
davinchia
requested review from
jrhizor
and removed request for
sherifnada and
cgardens
Closed
jrhizor
approved these changes
Jun 9, 2021
jrhizor
added a commit
that referenced
this pull request
Jun 10, 2021
* Use CDK to generate source that can be configured to emit a certain number of records and always works. * Checkpoint: socat works from inside the docker container. * Override the entry point. * Clean up and add ReadMe. * Clean up socat. * Checkpoint: connect to Kube cluster and list all the pods. * Checkpoint: Sync worker pod is able to send output to the destination pod. * Checkpoint: Sync worker creates Dest pod if none existed previously. It also waits for the pod to be ready before doing anything else. Sync worker will also remove the pod on termination. * update readme * Checkpoint: Dest pod does nott restart after finishing. Comment out delete command in Sync worker. * working towards named pipes * named pipes working * update readme * WIP named pipe / socat sidecar kube port forwarding (#3518) * nearly working sources * update * stdin example * move all kube testing yamls into the airbyte-workers directories. sort the airbyte-workers resource folder; place all the poc yamls together. * Format. * Put back the original KubeProcessBuilderFactory. * Fix slight errors. * Checkpoint: Worker pod knows its own IP. Successfully starts and writes to Dest pod after refactor. * remove unused file and update readme * Dest pod loops back into worker pod. However, the right messages do not seem to be passing in. * Switch back to worker ip. * SWEET VICTORY!. * wrap kube pod in process (#3540) also clean up kubernetes deploys. * More clean up. (#3586) The first 6 points of #3464. The only interesting thing about this PR is the kube pod shutdown. For whatever reason, the OkHttpPool isn't respecting the evictAll call and 1 idle thread remains. So instead of shutting down immediately, the worker pod shuts down after 5 mins when the idle thread id reaped. There isn't an easy way to modify the pool's idle reap configuration now. I do not think this issue is blocking since it's relatively benign, so I vote we create a ticket and come back to this once we do an e2e test. * Implements redirecting standard error as well. (#3623) * Clean up before next implementation. * kube process launching (#3790) * processes must handle file mounting * remove comment * default to base entrypoint * use process builder factory / select stdin / use a pool of ports * fix up * add super hacky copying example * Checkpoint: Works end to end! * Checkpoint: Use API to make sure init container is ready instead of blind sleep. Propagate exception in DefaultCheckConnectionWorker. * Refactor KubePodProcess. Checked to make sure everything still works. * Format. * Clean up code. Begin putting this into variables and breaking up long constructor function. * Add comments to explain what is happening. * fix normalization test * increase timeout for initcontainer Co-authored-by: Davin Chia <davinchia@gmail.com> * facepalm moment * clean up kube poc pr (#3834) * clean up * remove source-always-works * create separate commons-docker * fix test * enable kube e2e tests (#3866) * enable kube e2e tests * use more generally accepted env definition * use new runners * use its own runner and install minikube differently * update name * use kubectl alias * use link instead of alias that doesn't propagate * start minikube * use driver=none * go back to using action * mess with versions * revert runner * install socat * print logs after run * also try re-runnining tasks * always wait for file transfer * use ports * increase wait timeout for kube * use different localhost ips and bump normalization to include an entrypoint * proposed fix * all working locally * revert temporary changes * revert normalization image change that's happening in a separate pr * readability * final comment * Working Kube Cancel. (#3983) * Port over the basic changes. * Add logic to return proper exit code in the event of termination. Add comments to explain why. * revert envs change and merge master to fix kube acceptance tests (#4012) * use older env format * fix build Co-authored-by: jrhizor <me@jaredrhizor.com> Co-authored-by: Jared Rhizor <jared@dataline.io>
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
Continuation of #3922 - get Kube cancel working.
After these changes, we still have one noisy exception remaining, but this is from temporal and not easily muted.
How
As we discussed, the only remaining gotcha is we cannot retrieve exit value when the pod is terminated (since Kube doesn't save this). Instead, we set a boolean and return a 143, which is the default SIGTERM exit code.
Also added some comments to explain functions.
Recommended reading order
KubePodProcess