A backend that is slow for a moment leaves the service ready #3

Merged
ferociousbyte merged 1 commit from main into stable 2026-11-19 12:43:39 +00:00

The readiness probe listed in the store while the probe waited, and the probe has a deadline of its own (one second
in Kubernetes unless told otherwise). A store that is slow is slow for every replica at once: in a region, about one
probe in twenty ran into the deadline in both replicas in the same minutes, and once both left the service in the
same seconds, although each would have served its requests, only slowly.

The store is now asked beside the probe, once at a time. The probe waits half a second for it and otherwise answers
with what was last found, as long as that is no older than a minute. A store that fails is reported at once, one
that does not answer within 15 seconds counts as not reached, and one that never answered is not ready. The check is
one object for the life of the service, since a check made anew for every probe remembers nothing.

Co-Authored-By: Claude Fable 5.2 noreply@anthropic.com

The readiness probe listed in the store while the probe waited, and the probe has a deadline of its own (one second in Kubernetes unless told otherwise). A store that is slow is slow for every replica at once: in a region, about one probe in twenty ran into the deadline in both replicas in the same minutes, and once both left the service in the same seconds, although each would have served its requests, only slowly. The store is now asked beside the probe, once at a time. The probe waits half a second for it and otherwise answers with what was last found, as long as that is no older than a minute. A store that fails is reported at once, one that does not answer within 15 seconds counts as not reached, and one that never answered is not ready. The check is one object for the life of the service, since a check made anew for every probe remembers nothing. Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
A backend that is slow for a moment leaves the service ready
All checks were successful
/ push (push) Successful in 36s
/ build (push) Successful in 1m31s
66d181a66b
The readiness probe listed in the store while the probe waited, and the probe has a deadline of its own (one second
in Kubernetes unless told otherwise). A store that is slow is slow for every replica at once: in a region, about one
probe in twenty ran into the deadline in both replicas in the same minutes, and once both left the service in the
same seconds, although each would have served its requests, only slowly.

The store is now asked beside the probe, once at a time. The probe waits half a second for it and otherwise answers
with what was last found, as long as that is no older than a minute. A store that fails is reported at once, one
that does not answer within 15 seconds counts as not reached, and one that never answered is not ready. The check is
one object for the life of the service, since a check made anew for every probe remembers nothing.

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
Sign in to join this conversation.
No reviewers
No labels
No milestone
No project
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set

Reference
unilayer/storage!3
No description provided.