Add native memory circuit breaker. by jngz-es · Pull Request #689 · opensearch-project/ml-commons

jngz-es · 2023-01-11T00:48:46Z

Add native memory circuit breaker.
Refactor all breakers from common to plugin.
Add dynamic setting for native memory circuit breaker.

Signed-off-by: Jing Zhang jngz@amazon.com

Description

Add native memory circuit breaker.
Refactor all breakers from common to plugin.
Add dynamic setting for native memory circuit breaker.

Issues Resolved

#648

Check List

New functionality includes testing.
- All tests pass
New functionality has been documented.
- New functionality has javadoc added
Commits are signed per the DCO using --signoff

By submitting this pull request, I confirm that my contribution is made under the terms of the Apache 2.0 license.
For more information on following Developer Certificate of Origin and signing off your commits, please check here.

Refactor all breakers from common to plugin. Add dynamic setting for native memory circuit breaker. Signed-off-by: Jing Zhang <jngz@amazon.com>

codecov-commenter · 2023-01-11T00:57:14Z

Codecov Report

Merging #689 (fc43c48) into 2.x (9420c91) will decrease coverage by 0.41%.
The diff coverage is 85.29%.

@@             Coverage Diff              @@
##                2.x     #689      +/-   ##
============================================
- Coverage     83.91%   83.50%   -0.42%     
- Complexity     1008     1021      +13     
============================================
  Files            93       99       +6     
  Lines          3662     3740      +78     
  Branches        343      344       +1     
============================================
+ Hits           3073     3123      +50     
- Misses          444      470      +26     
- Partials        145      147       +2

Flag	Coverage Δ
ml-commons	`83.50% <85.29%> (-0.42%)`	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

Impacted Files	Coverage Δ
...ml/action/load/TransportLoadModelOnNodeAction.java	`97.29% <ø> (ø)`
.../org/opensearch/ml/breaker/DiskCircuitBreaker.java	`36.36% <0.00%> (ø)`
...rg/opensearch/ml/breaker/MemoryCircuitBreaker.java	`100.00% <ø> (ø)`
...opensearch/ml/breaker/ThresholdCircuitBreaker.java	`100.00% <ø> (ø)`
...n/java/org/opensearch/ml/model/MLModelManager.java	`78.20% <ø> (ø)`
...va/org/opensearch/ml/task/MLExecuteTaskRunner.java	`83.33% <ø> (ø)`
...va/org/opensearch/ml/task/MLPredictTaskRunner.java	`82.14% <ø> (ø)`
...main/java/org/opensearch/ml/task/MLTaskRunner.java	`97.72% <ø> (ø)`
...pensearch/ml/task/MLTrainAndPredictTaskRunner.java	`78.26% <ø> (ø)`
...a/org/opensearch/ml/task/MLTrainingTaskRunner.java	`74.76% <ø> (ø)`
... and 15 more

Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here.

ylwu-amzn · 2023-01-11T00:57:44Z

{"error":{"root_cause":[{"type":"illegal_argument_exception","reason":"persistent setting [plugins.ml_commons.native_mem_threshold], not recognized"}],"type":"illegal_argument_exception","reason":"persistent setting [plugins.ml_commons.native_mem_threshold], not recognized"},"status":400}

Sicheng added BWC test, which test from 2.4.0, in 2.4.0, we don't have the native memory threshold setting

ylwu-amzn · 2023-01-11T00:58:44Z

plugin/src/main/java/org/opensearch/ml/settings/MLCommonsSettings.java

        );
+
+    public static final Setting<Integer> ML_COMMONS_NATIVE_MEM_THRESHOLD = Setting
+        .intSetting("plugins.ml_commons.native_mem_threshold", 90, 0, 100, Setting.Property.NodeScope, Setting.Property.Dynamic);


mem seems not so clear. how about change to

Suggested change

.intSetting("plugins.ml_commons.native_mem_threshold", 90, 0, 100, Setting.Property.NodeScope, Setting.Property.Dynamic);

.intSetting("plugins.ml_commons.native_memory_threshold", 90, 0, 100, Setting.Property.NodeScope, Setting.Property.Dynamic);

sure, will change it to memory.

ylwu-amzn · 2023-01-11T01:12:19Z

plugin/src/main/java/org/opensearch/ml/breaker/DiskCircuitBreaker.java

        try {
            return AccessController.doPrivileged((PrivilegedExceptionAction<Boolean>) () -> {
-                return (new File(diskDir).getFreeSpace()/1024/1024/1024) < getThreshold();  // in GB
+                return (new File(diskDir).getFreeSpace() / 1024 / 1024 / 1024) < getThreshold();  // in GB


minor: we can create constant for the disk threshold rather than recalculate for every request.

public static final long DEFAULT_FREE_DISK_SPACE_THRESHOLD = 5l * 1024 * 1024 * 1024;

It's from original code not for this PR. Sure, I can add it.

I prefer to add a contant only for 1024 * 1024 * 1024, as the original logic is using G as default unit.

ylwu-amzn · 2023-01-11T01:19:06Z

plugin/src/main/java/org/opensearch/ml/breaker/MLCircuitBreakerService.java

        for (CircuitBreaker breaker : breakers.values()) {
            if (breaker.isOpen()) {
-                return breaker.getName();
+                return (ThresholdCircuitBreaker) breaker;


breaker class is CircuitBreaker, this line enforce it to ThresholdCircuitBreaker, seems not so extensible. If we have new child class of CircuitBreaker, this line will throw error.

Why need to change the return type from String to ThresholdCircuitBreaker?

Want to get more information like the current threshold when breaker is triggered to help debug.

I think we can add a warning log to help

Signed-off-by: Jing Zhang <jngz@amazon.com>

ylwu-amzn · 2023-01-11T02:09:21Z

plugin/src/main/java/org/opensearch/ml/breaker/DiskCircuitBreaker.java

        try {
            return AccessController.doPrivileged((PrivilegedExceptionAction<Boolean>) () -> {
-                return (new File(diskDir).getFreeSpace()/1024/1024/1024) < getThreshold();  // in GB
+                return (new File(diskDir).getFreeSpace() / GB) < getThreshold();  // in GB


Can we just set threshold and compare directly without dividing GB for every request? Just like what you did for native memory breaker?

Signed-off-by: Jing Zhang <jngz@amazon.com>

b4sjoo · 2023-01-11T01:14:22Z

plugin/src/test/java/org/opensearch/ml/model/MLModelManagerTests.java

+        when(thresholdCircuitBreaker.getThreshold()).thenReturn(87);
        expectedEx.expect(MLException.class);
-        expectedEx.expectMessage("Disk Circuit Breaker is open, please check your resources!");
+        expectedEx.expectMessage("Disk Circuit Breaker is open, threshold is 87. Please check your resources!");


Why use 87?

* Add native memory circuit breaker. Refactor all breakers from common to plugin. Add dynamic setting for native memory circuit breaker. Signed-off-by: Jing Zhang <jngz@amazon.com> * Address the comments 1. Signed-off-by: Jing Zhang <jngz@amazon.com> * Spotless changes. Signed-off-by: Jing Zhang <jngz@amazon.com> * Address the comments 2. Signed-off-by: Jing Zhang <jngz@amazon.com> Signed-off-by: Jing Zhang <jngz@amazon.com> (cherry picked from commit d809dd2)

* Add native memory circuit breaker. Refactor all breakers from common to plugin. Add dynamic setting for native memory circuit breaker. Signed-off-by: Jing Zhang <jngz@amazon.com> * Address the comments 1. Signed-off-by: Jing Zhang <jngz@amazon.com> * Spotless changes. Signed-off-by: Jing Zhang <jngz@amazon.com> * Address the comments 2. Signed-off-by: Jing Zhang <jngz@amazon.com> Signed-off-by: Jing Zhang <jngz@amazon.com> (cherry picked from commit d809dd2) Co-authored-by: Jing Zhang <jngz@amazon.com>

* Add native memory circuit breaker. Refactor all breakers from common to plugin. Add dynamic setting for native memory circuit breaker. Signed-off-by: Jing Zhang <jngz@amazon.com> * Address the comments 1. Signed-off-by: Jing Zhang <jngz@amazon.com> * Spotless changes. Signed-off-by: Jing Zhang <jngz@amazon.com> * Address the comments 2. Signed-off-by: Jing Zhang <jngz@amazon.com> Signed-off-by: Jing Zhang <jngz@amazon.com>

* Add native memory circuit breaker. Refactor all breakers from common to plugin. Add dynamic setting for native memory circuit breaker. * Address the comments 1. * Spotless changes. * Address the comments 2. Signed-off-by: Jing Zhang <jngz@amazon.com> Co-authored-by: Jing Zhang <jngz@amazon.com>

Add native memory circuit breaker.

eb03a32

Refactor all breakers from common to plugin. Add dynamic setting for native memory circuit breaker. Signed-off-by: Jing Zhang <jngz@amazon.com>

jngz-es requested a review from a team January 11, 2023 00:48

ylwu-amzn reviewed Jan 11, 2023

View reviewed changes

jngz-es added 2 commits January 10, 2023 17:43

Address the comments 1.

6493275

Signed-off-by: Jing Zhang <jngz@amazon.com>

Spotless changes.

fc43c48

Signed-off-by: Jing Zhang <jngz@amazon.com>

jngz-es requested a review from ylwu-amzn January 11, 2023 01:51

ylwu-amzn reviewed Jan 11, 2023

View reviewed changes

Address the comments 2.

8baa08f

Signed-off-by: Jing Zhang <jngz@amazon.com>

jngz-es requested a review from ylwu-amzn January 11, 2023 02:29

ylwu-amzn approved these changes Jan 11, 2023

View reviewed changes

ylwu-amzn added feature backport 2.5 labels Jan 11, 2023

b4sjoo approved these changes Jan 11, 2023

View reviewed changes

jngz-es merged commit d809dd2 into opensearch-project:2.x Jan 11, 2023

opensearch-trigger-bot bot mentioned this pull request Jan 11, 2023

[Backport 2.5] Add native memory circuit breaker. #693

Merged

ylwu-amzn mentioned this pull request Jan 11, 2023

NativeMemoryCircuitBreaker #675

Closed

5 tasks

ylwu-amzn mentioned this pull request Mar 2, 2023

[Backport to main] Add native memory circuit breaker. (#689) #777

Merged

5 tasks

jngz-es deleted the native branch May 23, 2023 20:45

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Add native memory circuit breaker.#689

Add native memory circuit breaker.#689
jngz-es merged 4 commits intoopensearch-project:2.xfrom
jngz-es:native

jngz-es commented Jan 11, 2023

Uh oh!

codecov-commenter commented Jan 11, 2023 •

edited

Loading

Uh oh!

ylwu-amzn commented Jan 11, 2023

Uh oh!

ylwu-amzn Jan 11, 2023

Uh oh!

jngz-es Jan 11, 2023

Uh oh!

ylwu-amzn Jan 11, 2023

Uh oh!

jngz-es Jan 11, 2023

Uh oh!

jngz-es Jan 11, 2023

Uh oh!

ylwu-amzn Jan 11, 2023

Uh oh!

jngz-es Jan 11, 2023

Uh oh!

ylwu-amzn Jan 11, 2023

Uh oh!

ylwu-amzn Jan 11, 2023

Uh oh!

b4sjoo Jan 11, 2023

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

	.intSetting("plugins.ml_commons.native_mem_threshold", 90, 0, 100, Setting.Property.NodeScope, Setting.Property.Dynamic);
	.intSetting("plugins.ml_commons.native_memory_threshold", 90, 0, 100, Setting.Property.NodeScope, Setting.Property.Dynamic);

Conversation

jngz-es commented Jan 11, 2023

Description

Issues Resolved

Check List

Uh oh!

codecov-commenter commented Jan 11, 2023 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codecov Report

Uh oh!

ylwu-amzn commented Jan 11, 2023

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

codecov-commenter commented Jan 11, 2023 •

edited

Loading