Stats rework (1/3): Interfaces and implementations for tiers by peteralfonsi · Pull Request #24 · peteralfonsi/OpenSearch

peteralfonsi · 2024-02-13T22:16:19Z

Description

Changes stats for ICache to be tracked by ICache implementations themselves, using the CacheStats object. ICache<K, V> implementations now take in ICacheKey, which contains a K as well as a list of CacheStatsDimension objects. This contains information that can be used to aggregate cache stats. For example, you might want to aggregate by shard ID value. CacheStats can return values either in total or using a list of dimensions.

Also modifies EhcacheDiskCache and OpenSearchOnHeapCache to use the changes to ICache, and to use the new SingleDimensionCacheStats objects to track their stats.

Related Issues

Part of tiered caching (stats rework)

Check List

New functionality includes testing.
- All tests pass
New functionality has been documented.
- New functionality has javadoc added
Failing checks are inspected and point to the corresponding known issue(s) (See: Troubleshooting Failing Builds)
Commits are signed per the DCO using --signoff
Commit changes are listed out in CHANGELOG.md file (See: Changelog)
Public documentation issue/PR created

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.

Signed-off-by: Peter Alfonsi <petealft@amazon.com>

…cheStatsDimension Signed-off-by: Peter Alfonsi <petealft@amazon.com>

Signed-off-by: Peter Alfonsi <petealft@amazon.com>

…te stats Signed-off-by: Peter Alfonsi <petealft@amazon.com>

…ended on it Signed-off-by: Peter Alfonsi <petealft@amazon.com>

Signed-off-by: Peter Alfonsi <petealft@amazon.com>

sgup432 · 2024-02-26T17:33:51Z

server/src/main/java/org/opensearch/common/cache/stats/CacheStatsDimension.java

+    public static final String TIER_DIMENSION_VALUE_ON_HEAP = "on_heap";
+    public static final String TIER_DIMENSION_VALUE_DISK = "disk";


This should not be declared here. Should be present in respective caches.

sgup432 · 2024-02-26T17:41:36Z

server/src/main/java/org/opensearch/common/cache/store/OpenSearchOnHeapCache.java

        }
        cache = cacheBuilder.build();
-        this.eventListener = builder.getEventListener();
+        String dimensionName = Objects.requireNonNull(builder.shardIdDimensionName, "Shard id dimension name can't be null");


This cache shouldn't be limited to a specific dimension. It is used by IRC this way but we need to make it generic(in terms of dimension) going forward. Same with other caches.

sgup432 · 2024-02-26T17:41:38Z

server/src/main/java/org/opensearch/common/cache/store/OpenSearchOnHeapCache.java

        cache = cacheBuilder.build();
-        this.eventListener = builder.getEventListener();
+        String dimensionName = Objects.requireNonNull(builder.shardIdDimensionName, "Shard id dimension name can't be null");
+        this.stats = new SingleDimensionCacheStats(dimensionName, CacheStatsDimension.TIER_DIMENSION_VALUE_ON_HEAP);


This doesn't look right. Why are we limiting this to a single dimension?

sgup432 · 2024-02-26T17:43:54Z

server/src/main/java/org/opensearch/common/cache/stats/SingleDimensionCacheStats.java

+ * A CacheStats implementation for caches that aggregate over a single dimension, as well as holding a tier dimension.
+ * For example, caches in the IndicesRequestCache only aggregate over ShardId value.
+ */
+public class SingleDimensionCacheStats implements CacheStats {


Lets not do this. In case I want to have multiple dimension, I would have to implement another class?
EhcacheDiskCache or onHeap cache should have the capability to support multiple dimensions from the start and keep it generic.

You can implement this support multiple dimensions and it can be used to have a single dimension if needed. That makes it extensible.

sgup432 · 2024-02-26T17:44:52Z

plugins/cache-ehcache/src/main/java/org/opensearch/cache/store/disk/EhcacheDiskCache.java

+            Objects.requireNonNull(builder.getWeigher(), "Weigher function can't be null"),
+            this.valueSerializer);
        this.cache = buildCache(Duration.ofMillis(expireAfterAccess.getMillis()), builder);
+        this.shardIdDimensionName = Objects.requireNonNull(builder.shardIdDimensionName, "Dimension name can't be null");


This is not right. We shouldn't be limiting this to a single dimension that too hardcoding it to be shardId specific.

…ed-stats-tiers

Signed-off-by: Peter Alfonsi <petealft@amazon.com>

Signed-off-by: Sagar Upadhyaya <sagar.upadhyaya.121@gmail.com>

Signed-off-by: Peter Alfonsi <petealft@amazon.com>

…ed-stats-tiers

Signed-off-by: Peter Alfonsi <petealft@amazon.com>

Peter Alfonsi added 17 commits February 8, 2024 11:36

CacheStats interface and SingleDimensionCacheStats impl

641abf1

Signed-off-by: Peter Alfonsi <petealft@amazon.com>

First attempt to integrate new stats with EhcacheDiskCache

937ca51

Signed-off-by: Peter Alfonsi <petealft@amazon.com>

Adds Serializer interface and impls for Key and BytesReference

21b1078

Signed-off-by: Peter Alfonsi <petealft@amazon.com>

Adds ICacheKey serializer impl

a460acd

Signed-off-by: Peter Alfonsi <petealft@amazon.com>

Added more icachekey serializer tests

31ee9ae

Signed-off-by: Peter Alfonsi <petealft@amazon.com>

Attempts to fix ehcache hits

9069077

Signed-off-by: Peter Alfonsi <petealft@amazon.com>

Fixed unexpected misses by implementing hashCode for ICacheKey and Ca…

18ddb23

…cheStatsDimension Signed-off-by: Peter Alfonsi <petealft@amazon.com>

Adds memory tracking to disk tier

8618458

Signed-off-by: Peter Alfonsi <petealft@amazon.com>

Updated other tests

46b3e49

Signed-off-by: Peter Alfonsi <petealft@amazon.com>

Added partial memory tests

51ea583

Signed-off-by: Peter Alfonsi <petealft@amazon.com>

Changed cache stats API to get values by dimension list

17cf26a

Signed-off-by: Peter Alfonsi <petealft@amazon.com>

Cleanup

be5eece

Signed-off-by: Peter Alfonsi <petealft@amazon.com>

Redid memory tracking to use already implemented weigher fn

a6b0899

Signed-off-by: Peter Alfonsi <petealft@amazon.com>

Split CacheStats into CacheStats and CacheStatsBase, which can't upda…

9b1e66c

…te stats Signed-off-by: Peter Alfonsi <petealft@amazon.com>

Readded BytesReferenceSerializer impl as ICacheKeySerializerTests dep…

01e04a8

…ended on it Signed-off-by: Peter Alfonsi <petealft@amazon.com>

Changed SingleDimensionCacheStats to use ConcurrentMap

3777e3f

Signed-off-by: Peter Alfonsi <petealft@amazon.com>

Made SingleDimensionCacheStats also take in tier dimensions

6059680

Signed-off-by: Peter Alfonsi <petealft@amazon.com>

peteralfonsi force-pushed the reworked-stats-tiers branch from 67e68c4 to 6059680 Compare February 14, 2024 23:04

Peter Alfonsi added 2 commits February 14, 2024 15:51

Added overall CacheStatsResponse object packaging all 5 metrics

7f5a455

Signed-off-by: Peter Alfonsi <petealft@amazon.com>

Removed skeleton TSC stats implementation

36c600d

Signed-off-by: Peter Alfonsi <petealft@amazon.com>

peteralfonsi force-pushed the reworked-stats-tiers branch from e28c01d to 36c600d Compare February 14, 2024 23:53

sgup432 reviewed Feb 26, 2024

View reviewed changes

Peter Alfonsi and others added 8 commits February 26, 2024 12:43

Merge remote-tracking branch 'sgup432/ehcache_disk_integ' into rework…

27fdfe1

…ed-stats-tiers

Modified factories to take new arguments

c8dc1b3

Signed-off-by: Peter Alfonsi <petealft@amazon.com>

added utilitly fns to CacheStatsResponse

4abd602

Signed-off-by: Peter Alfonsi <petealft@amazon.com>

Making TieredCachePlugin constructor public

2586fa1

Signed-off-by: Sagar Upadhyaya <sagar.upadhyaya.121@gmail.com>

Fixing CacheService unit test

6a2b374

Signed-off-by: Sagar Upadhyaya <sagar.upadhyaya.121@gmail.com>

Added multi dimension cache stats impl

3e7ea26

Signed-off-by: Peter Alfonsi <petealft@amazon.com>

Removed SingleDimensionCacheStats

b4c83e7

Signed-off-by: Peter Alfonsi <petealft@amazon.com>

Adds IRC key serializer

d579c51

Signed-off-by: Peter Alfonsi <petealft@amazon.com>

Peter Alfonsi added 7 commits February 27, 2024 13:54

Addressed Sagar's other comment

c34c218

Signed-off-by: Peter Alfonsi <petealft@amazon.com>

Merge remote-tracking branch 'sgup432/ehcache_disk_integ' into rework…

2f4bd4f

…ed-stats-tiers

Fixed on heap stats integration, added tests

a61f033

Signed-off-by: Peter Alfonsi <petealft@amazon.com>

Optimized multi dimension stats

2483981

Signed-off-by: Peter Alfonsi <petealft@amazon.com>

Added reset() to CacheStats

2aeaa53

Signed-off-by: Peter Alfonsi <petealft@amazon.com>

Fixed reset impl for multi dim stats

60df761

Signed-off-by: Peter Alfonsi <petealft@amazon.com>

spotlessApply

ea33af8

Signed-off-by: Peter Alfonsi <petealft@amazon.com>

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Stats rework (1/3): Interfaces and implementations for tiers#24

Stats rework (1/3): Interfaces and implementations for tiers#24
peteralfonsi wants to merge 34 commits intoehcache-disk-integ-basefrom
reworked-stats-tiers

peteralfonsi commented Feb 13, 2024

Uh oh!

sgup432 Feb 26, 2024

Uh oh!

sgup432 Feb 26, 2024

Uh oh!

sgup432 Feb 26, 2024

Uh oh!

sgup432 Feb 26, 2024

Uh oh!

sgup432 Feb 26, 2024

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

		public static final String TIER_DIMENSION_VALUE_ON_HEAP = "on_heap";
		public static final String TIER_DIMENSION_VALUE_DISK = "disk";

Conversation

peteralfonsi commented Feb 13, 2024

Description

Related Issues

Check List

Uh oh!

sgup432 Feb 26, 2024

Choose a reason for hiding this comment

Uh oh!

sgup432 Feb 26, 2024

Choose a reason for hiding this comment

Uh oh!

sgup432 Feb 26, 2024

Choose a reason for hiding this comment

Uh oh!

sgup432 Feb 26, 2024

Choose a reason for hiding this comment

Uh oh!

sgup432 Feb 26, 2024

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants