- Notifications
You must be signed in to change notification settings - Fork 25.5k
[ML] Set Adaptive Allocations Scale to Zero to 24h #128914
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
Adaptive Allocations will now wait 24 hours to scale down to zero allocations (from one allocation). `xpack.ml.trained_models.adaptive_allocations.scale_to_zero_time` can be used to change this time from 24 hours to a minimum of 15 minutes. `xpack.ml.trained_models.adaptive_allocations.scale_cooldown_time` can be used to change the time adaptive allocation scales between 1 and max allocations. Defaults to 5 minutes. `xpack.ml.trained_models.adaptive_allocations.trigger_time` can be used to change the time adaptive allocations tries to scale. Defaults to 15 minutes.
Hi @prwhelan, I've created a changelog YAML for you. |
Pinging @elastic/ml-core (Team:ML) |
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
docs/changelog/128914.yaml Outdated
@@ -0,0 +1,5 @@ | |||
pr: 128914 | |||
summary: Set Adaptive Allocations Scale to Zero to 24h |
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.
should be: "make configurable ... and set default to 24h"
public static final Setting<TimeValue> DEFAULT_TIME_INTERVAL = Setting.timeSetting( | ||
"xpack.ml.trained_models.adaptive_allocations.trigger_time", | ||
TimeValue.timeValueSeconds(10), | ||
TimeValue.timeValueSeconds(10), |
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.
Why the minimum times (here and underneath)?
I you want an allocation to go away after 1 minute of idleness, I think you should be able to do so
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.
I picked the current settings as the min settings, but we can lower them to 1 minute, I don't mind either way. I think values lower than 1m may cause too much thrashing for the autoscaler
public static final Setting<TimeValue> DEFAULT_TIME_INTERVAL = Setting.timeSetting( | ||
"xpack.ml.trained_models.adaptive_allocations.trigger_time", | ||
TimeValue.timeValueSeconds(10), | ||
TimeValue.timeValueSeconds(10), |
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.
BTW, I'm not convinced this trigger time should be configurable
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
Adaptive Allocations will now wait 24 hours to scale down to zero allocations (from one allocation). `xpack.ml.trained_models.adaptive_allocations.scale_to_zero_time` can be used to change this time from 24 hours to a minimum of 1 minute. `xpack.ml.trained_models.adaptive_allocations.scale_cooldown_time` can be used to change the time adaptive allocation scales between 1 and max allocations. Defaults to 5 minutes.
Adaptive Allocations will now wait 24 hours to scale down to zero allocations (from one allocation). `xpack.ml.trained_models.adaptive_allocations.scale_to_zero_time` can be used to change this time from 24 hours to a minimum of 1 minute. `xpack.ml.trained_models.adaptive_allocations.scale_cooldown_time` can be used to change the time adaptive allocation scales between 1 and max allocations. Defaults to 5 minutes.
Adaptive Allocations will now wait 24 hours to scale down to zero allocations (from one allocation).
xpack.ml.trained_models.adaptive_allocations.scale_to_zero_time
can be used to change this time from 24 hours to a minimum of 15 minutes.xpack.ml.trained_models.adaptive_allocations.scale_cooldown_time
can be used to change the time adaptive allocation scales between 1 and max allocations. Defaults to 5 minutes.