You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
This repository has been archived by the owner on Apr 27, 2023. It is now read-only.
We want to keep file uploads alive during rollouts of submission for a certain grace period.
This will partially rely on the k8s feature of preStop hooks. These delay the sending of SIGTERM until the hook(s) have completed. The hooks can be commands executed inside the container or http requests.
Note that if the overall grace period is exceeded during the preStop hook the container will receive a SIGKILL.
To avoid waiting needlessly, say by using sleep, we ideally need an http endpoint that answers with 200 once all or at least the upload connections have been closed.
This is a requirement for #998 and a precursor to the validation on submission in #1306
The text was updated successfully, but these errors were encountered:
After careful consideration, it made more sense to gracefully handle sigterm and wait for the close event to be handled. As reflected in libero/reviewer-submission#223
We want to keep file uploads alive during rollouts of submission for a certain grace period.
This will partially rely on the k8s feature of
preStop
hooks. These delay the sending of SIGTERM until the hook(s) have completed. The hooks can be commands executed inside the container or http requests.Note that if the overall grace period is exceeded during the preStop hook the container will receive a SIGKILL.
To avoid waiting needlessly, say by using
sleep
, we ideally need an http endpoint that answers with 200 once all or at least the upload connections have been closed.This is a requirement for #998 and a precursor to the validation on submission in #1306
The text was updated successfully, but these errors were encountered: