feat: enable draining csb via os signal handling #1096
Closed
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
When running as an app in CF we can rely on the platform to handle TLS setup, but on a VM currently there is no way to have encrypted traffic. TPCF-26820
it is angry about `.` imports. But we do not mind because this is not production code.
when a csb app that is running in cf is stopped outside of it's own lifecycle ( e.g. the diego cell is redeployed) we do not have a great way of ensuring that all in flight terraform executions will be able to finish their work and write back the resulting tf state to the csb DB. Diego assumes that an app will gracefully shutdown within 10s of receiving SIGTERM, if that is not the case, the App will receive a SIGKILL and stop abruptly That creates orphaned resources in the underlying IaaS that cannot be cleaned up by the csb because the CSB does not have the tfstate for the terraform resources that were in flight when the CSB got shutdown. To aleviate this issue, this introduces a graceful shutdown sequence and and lockfiles on disk ( to be consumed by a drain script ). This enables to deploy the CSB as a workload on a bosh instance. Instead of marking specific SI instances as failed, this ensures that the broker will - stop accepting new requests - finish all in flight TF before shutdown. The drain script can be kept simple by inspecting a folder. If that folder is empty, it is safe to proceed to stop the CSB. We also tried a drain script based on inspecting the processes running ( e.g. if a tofu or provider binary is still being executed ). Though that seems potentially unreliable ( since there could be time of check // time of use issues ) that falsely suggest that everything is finished ( e.g. because we checked right between two invocations of the provider / tofu binaries ) - fly-by: some structs got their fiels reordered to improve their memory footprint.
* extended inflight operation test to check deprovision * removed focus test * removed failing check for SIGTERM, in this case SIGKILL is sent - but not seen in log
A previous commit fixed an issue with LockFilesExist returning an inverted value. The existing drain wait code depended on this incorrect behaviour.
I O U a proper message
FelisiaM
changed the title
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.
Checklist: