kanaloa.reactive.dispatcher.queue
AutoScalingSettings
Related Doc:
package queue
chance of scaling down when the worker pool is fully utlized
duration between each scaling attempt
start to downsize after underutilized for period, should be long enough to include at least one traffic cycle.
during optimization, it only looks at this number of adjacent pool sizes (adjacent to current pool size), to figure out the optimal pool size to move to
during exploration, it takes as big a step as this size. It's a ratio to the current pool size, so if the current size is 10 and the exploreStepSize is 0.2, the exploration will be within a range between 8 and 12
during downsizing, it will downsize to the largest number of concurrently occupied workers it has seen plus a buffer zone, this downsizeRatio determines the buffer size.
chance of doing a exploration vs an optimization
maximum time allowed when autoscaler collects status from queue, queueProcessor and all workers
duration between each scaling attempt
chance of scaling down when the worker pool is fully utlized
start to downsize after underutilized for period, should be long enough to include at least one traffic cycle.
during downsizing, it will downsize to the largest number of concurrently occupied workers it has seen plus a buffer zone, this downsizeRatio determines the buffer size.
chance of doing a exploration vs an optimization
during exploration, it takes as big a step as this size.
during exploration, it takes as big a step as this size. It's a ratio to the current pool size, so if the current size is 10 and the exploreStepSize is 0.2, the exploration will be within a range between 8 and 12
during optimization, it only looks at this number of adjacent pool sizes (adjacent to current pool size), to figure out the optimal pool size to move to
maximum time allowed when autoscaler collects status from queue, queueProcessor and all workers
chance of scaling down when the worker pool is fully utlized
duration between each scaling attempt
start to downsize after underutilized for period, should be long enough to include at least one traffic cycle.
during optimization, it only looks at this number of adjacent pool sizes (adjacent to current pool size), to figure out the optimal pool size to move to
during exploration, it takes as big a step as this size. It's a ratio to the current pool size, so if the current size is 10 and the exploreStepSize is 0.2, the exploration will be within a range between 8 and 12
during downsizing, it will downsize to the largest number of concurrently occupied workers it has seen plus a buffer zone, this downsizeRatio determines the buffer size.
chance of doing a exploration vs an optimization
maximum time allowed when autoscaler collects status from queue, queueProcessor and all workers