-
Notifications
You must be signed in to change notification settings - Fork 25.6k
adaptive allocations: reset time interval with zero requests #115400
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Conversation
…rting an allocation
Pinging @elastic/ml-core (Team:ML) |
|
||
AdaptiveAllocationsScaler scaler = scalers.get(assignment.getDeploymentId()); | ||
if (scaler != null) { | ||
scaler.resetTimeWithoutRequests(); |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
If starting an allocation takes too long (longer than the inference request timeout of 10secs), the inference request fails prematurely, and never reaches the allocation. This means that it isn't counted in the node stats, the "timeWithoutRequests" of the autoscaler keeps ticking, and the new allocation is stopped too soon.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
LGTM
💔 Backport failedThe backport operation could not be completed due to the following error:
You can use sqren/backport to manually backport by running |
…rting an allocation (#115400)
…rting an allocation (#115400)
…rting an allocation (elastic#115400)
…rting an allocation (elastic#115400)
…rting an allocation (elastic#115400)
No description provided.