IGNITE-29032 Improvement of the page eviction mechanism for in-memory by wernerdv · Pull Request #13554 · apache/ignite

wernerdv · 2026-09-04T08:25:44Z

Thank you for submitting the pull request to the Apache Ignite.

In order to streamline the review of the contribution
we ask you to ensure the following steps have been taken:

The Contribution Checklist

There is a single JIRA ticket related to the pull request.
The web-link to the pull request is attached to the JIRA ticket.
The JIRA ticket has the Patch Available state.
The pull request body describes changes that have been made.
The description explains WHAT and WHY was made instead of HOW.
The pull request title is treated as the final commit message.
The following pattern must be used: IGNITE-XXXX Change summary where XXXX - number of JIRA issue.
A reviewer has been mentioned through the JIRA comments
(see the Maintainers list)
The pull request has been checked by the Teamcity Bot and
the green visa attached to the JIRA ticket (see tab PR Check at TC.Bot - Instance 1 or TC.Bot - Instance 2)

Notes

If you need any help, please email dev@ignite.apache.org or ask anу advice on http://asf.slack.com #ignite channel.

ignitetcbot · 2026-09-04T12:49:29Z

TCBot Test Analysis

TeamCity: --> Run :: All Results
Branch: pull/13554/head
Base: master
Tested commit: ee910c0

Possible Blockers (0)

No blockers found.

New Tests (18)

Cache 21: 18 tests
- IgniteCacheTestSuite19: Random2LruPageEvictionSizeAwareTest.testPutAllLargeRows - PASSED
- IgniteCacheTestSuite19: Random2LruPageEvictionSizeAwareTest.testUpdateRowGrows - PASSED
- IgniteCacheTestSuite19: Random2LruPageEvictionSizeAwareTest.testLargeObjectReadBack - PASSED
- IgniteCacheTestSuite19: Random2LruPageEvictionSizeAwareTest.testPutLargeObjectsDoesNotOom - PASSED
- IgniteCacheTestSuite19: RandomLruPageEvictionWithExpiryPolicyTest.testLargePutWithExpiryNoDeadlock - PASSED
- IgniteCacheTestSuite19: RandomLruPageEvictionWithExpiryPolicyTest.testTtlFreedSpaceAccountedForByEviction - PASSED
- IgniteCacheTestSuite19: Random2LruPageEvictionWithExpiryPolicyTest.testLargePutWithExpiryNoDeadlock - PASSED
- IgniteCacheTestSuite19: Random2LruPageEvictionWithExpiryPolicyTest.testTtlFreedSpaceAccountedForByEviction - PASSED
- IgniteCacheTestSuite19: RandomLruPageEvictionSizeAwareTest.testUpdateRowGrows - PASSED
- IgniteCacheTestSuite19: RandomLruPageEvictionSizeAwareTest.testRecordLargerThanRegionOom - PASSED
- ... and 8 more new tests

Copilot

🟡 Changes recommended

One or more issues must be addressed before approval.

Once you've addressed the issues Copilot identified, you can request another Copilot review.

Pull request overview

Improves in-memory page eviction to support large rows and avoid lock-ordering deadlocks.

Changes:

Adds size-aware eviction for single and batch inserts.
Introduces non-blocking eviction locking and progress guards.
Expands eviction regression tests for both LRU modes.

File summaries

File	Description
modules/core/src/test/java/org/apache/ignite/testsuites/IgniteCacheEvictionSelfTestSuite.java	Updated as part of this pull request.
modules/core/src/test/java/org/apache/ignite/internal/processors/cache/eviction/paged/RandomLruPageEvictionWithExpiryPolicyTest.java	Updated as part of this pull request.
modules/core/src/test/java/org/apache/ignite/internal/processors/cache/eviction/paged/RandomLruPageEvictionSizeAwareTest.java	Updated as part of this pull request.
modules/core/src/test/java/org/apache/ignite/internal/processors/cache/eviction/paged/RandomLruPageEvictionConcurrentWritesTest.java	Updated as part of this pull request.
modules/core/src/test/java/org/apache/ignite/internal/processors/cache/eviction/paged/Random2LruPageEvictionWithExpiryPolicyTest.java	Updated as part of this pull request.
modules/core/src/test/java/org/apache/ignite/internal/processors/cache/eviction/paged/Random2LruPageEvictionSizeAwareTest.java	Updated as part of this pull request.
modules/core/src/test/java/org/apache/ignite/internal/processors/cache/eviction/paged/Random2LruPageEvictionConcurrentWritesTest.java	Updated as part of this pull request.
modules/core/src/test/java/org/apache/ignite/internal/processors/cache/eviction/paged/PageEvictionWithExpiryPolicyAbstractTest.java	Updated as part of this pull request.
modules/core/src/test/java/org/apache/ignite/internal/processors/cache/eviction/paged/PageEvictionSizeAwareAbstractTest.java	Updated as part of this pull request.
modules/core/src/test/java/org/apache/ignite/internal/processors/cache/eviction/paged/PageEvictionPutLargeObjectsAbstractTest.java	Updated as part of this pull request.
modules/core/src/test/java/org/apache/ignite/internal/processors/cache/eviction/paged/PageEvictionMetricTest.java	Updated as part of this pull request.
modules/core/src/test/java/org/apache/ignite/internal/processors/cache/eviction/paged/PageEvictionGuardOomTest.java	Updated as part of this pull request.
modules/core/src/test/java/org/apache/ignite/internal/processors/cache/eviction/paged/PageEvictionConcurrentWritesAbstractTest.java	Updated as part of this pull request.
modules/core/src/main/java/org/apache/ignite/internal/processors/cache/persistence/RowStore.java	Updated as part of this pull request.
modules/core/src/main/java/org/apache/ignite/internal/processors/cache/persistence/IgniteCacheDatabaseSharedManager.java	Updated as part of this pull request.
modules/core/src/main/java/org/apache/ignite/internal/processors/cache/persistence/evict/PageAbstractEvictionTracker.java	Updated as part of this pull request.
modules/core/src/main/java/org/apache/ignite/internal/processors/cache/GridCacheMapEntry.java	Updated as part of this pull request.
modules/core/src/main/java/org/apache/ignite/internal/processors/cache/GridCacheEntryEx.java	Updated as part of this pull request.
modules/core/src/main/java/org/apache/ignite/configuration/DataRegionConfiguration.java	Updated as part of this pull request.

Review details

Files reviewed: 19/19 changed files
Comments generated: 11
Review effort level: Balanced

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

ignitetcbot · 2026-09-08T07:03:22Z

TCBot Test Analysis

TeamCity: --> Run :: All Results
Branch: pull/13554/head
Base: master
Tested commit: 622b749

Possible Blockers (0)

No blockers found.

New Tests (18)

Cache 21: 18 tests
- IgniteCacheTestSuite19: Random2LruPageEvictionSizeAwareTest.testPutAllLargeRows - PASSED
- IgniteCacheTestSuite19: Random2LruPageEvictionSizeAwareTest.testUpdateRowGrows - PASSED
- IgniteCacheTestSuite19: Random2LruPageEvictionSizeAwareTest.testLargeObjectReadBack - PASSED
- IgniteCacheTestSuite19: Random2LruPageEvictionSizeAwareTest.testPutLargeObjectsDoesNotOom - PASSED
- IgniteCacheTestSuite19: RandomLruPageEvictionWithExpiryPolicyTest.testLargePutWithExpiryNoDeadlock - PASSED
- IgniteCacheTestSuite19: RandomLruPageEvictionWithExpiryPolicyTest.testTtlFreedSpaceAccountedForByEviction - PASSED
- IgniteCacheTestSuite19: Random2LruPageEvictionWithExpiryPolicyTest.testLargePutWithExpiryNoDeadlock - PASSED
- IgniteCacheTestSuite19: Random2LruPageEvictionWithExpiryPolicyTest.testTtlFreedSpaceAccountedForByEviction - PASSED
- IgniteCacheTestSuite19: RandomLruPageEvictionSizeAwareTest.testUpdateRowGrows - PASSED
- IgniteCacheTestSuite19: RandomLruPageEvictionSizeAwareTest.testRecordLargerThanRegionOom - PASSED
- ... and 8 more new tests

ignitetcbot · 2026-09-09T08:39:10Z

TCBot Test Analysis

TeamCity: --> Run :: All Results
Branch: pull/13554/head
Base: master
Tested commit: fb5b61e

Possible Blockers (0)

No blockers found.

New Tests (18)

Cache 21: 18 tests
- IgniteCacheTestSuite19: Random2LruPageEvictionSizeAwareTest.testPutAllLargeRows - PASSED
- IgniteCacheTestSuite19: Random2LruPageEvictionSizeAwareTest.testUpdateRowGrows - PASSED
- IgniteCacheTestSuite19: Random2LruPageEvictionSizeAwareTest.testLargeObjectReadBack - PASSED
- IgniteCacheTestSuite19: Random2LruPageEvictionSizeAwareTest.testPutLargeObjectsDoesNotOom - PASSED
- IgniteCacheTestSuite19: RandomLruPageEvictionWithExpiryPolicyTest.testLargePutWithExpiryNoDeadlock - PASSED
- IgniteCacheTestSuite19: RandomLruPageEvictionWithExpiryPolicyTest.testTtlFreedSpaceAccountedForByEviction - PASSED
- IgniteCacheTestSuite19: Random2LruPageEvictionWithExpiryPolicyTest.testLargePutWithExpiryNoDeadlock - PASSED
- IgniteCacheTestSuite19: Random2LruPageEvictionWithExpiryPolicyTest.testTtlFreedSpaceAccountedForByEviction - PASSED
- IgniteCacheTestSuite19: RandomLruPageEvictionSizeAwareTest.testUpdateRowGrows - PASSED
- IgniteCacheTestSuite19: RandomLruPageEvictionSizeAwareTest.testRecordLargerThanRegionOom - PASSED
- ... and 8 more new tests

wernerdv · 2026-09-09T09:29:33Z

JmhPageEvictionBenchmark results on my local machine:

master:

Benchmark                          (evictionMode)  (scenario)   Mode  Cnt    Score     Error   Units
JmhPageEvictionBenchmark.putLarge      RANDOM_LRU       SMALL  thrpt    5    0,733 ±   0,008  ops/ms
JmhPageEvictionBenchmark.putLarge      RANDOM_LRU       LARGE  thrpt    5    fail with IgniteOutOfMemoryException: Out of memory in data region
JmhPageEvictionBenchmark.putSmall      RANDOM_LRU       SMALL  thrpt    5  938,566 ±  70,609  ops/ms
JmhPageEvictionBenchmark.putSmall      RANDOM_LRU       LARGE  thrpt    5  820,833 ± 145,542  ops/ms

branch:

Benchmark                          (evictionMode)  (scenario)   Mode  Cnt    Score    Error   Units
JmhPageEvictionBenchmark.putLarge      RANDOM_LRU       SMALL  thrpt    5    0,789 ±  0,035  ops/ms
JmhPageEvictionBenchmark.putLarge      RANDOM_LRU       LARGE  thrpt    5    0,674 ±  0,018  ops/ms
JmhPageEvictionBenchmark.putSmall      RANDOM_LRU       SMALL  thrpt    5  944,594 ± 54,536  ops/ms
JmhPageEvictionBenchmark.putSmall      RANDOM_LRU       LARGE  thrpt    5  858,830 ± 37,303  ops/ms

alex-plekhanov · 2026-09-10T15:20:06Z

+     * while the calling thread already holds entry locks. When set, entries whose locks are contended are skipped
+     * (via a non-blocking {@code evictInternal}) instead of blocking, avoiding a lock-ordering deadlock.
+     */
+    private static final ThreadLocal<Boolean> EVICT_NON_BLOCKING = new ThreadLocal<>();


It's strange to use thread-local to pass parameter to next stack level. Let's use
evictDataPage(boolean blocking) -> evictDataPage(int pageIdx, boolean blocking)
Instead of
evictDataPageNonBlocking -> set thread-local -> evictDataPage() -> evictDataPage(int pageIdx) -> get thread local.

Agreed.
Moved the capability into the interface as evictDataPage(boolean tryLock) (default evictDataPage() delegates to false), removed the ThreadLocal.
As a follow-up the method now returns boolean so the size-aware progress guard can count an actual eviction as progress. Parameter named tryLock (matching evictInternal(..., tryLock) / lockEntry(boolean tryLock)) rather than blocking.

alex-plekhanov · 2026-09-10T15:31:01Z

+        // as a raw IgniteOutOfMemoryException (wrapped into CorruptedFreeListException in the batch path).
+        //
+        // The re-reserve is an inline demand-eviction: reached from the BPlusTree.invoke row-creation closure, it may
+        // re-entrantly remove other entries from the same data tree. That is safe because the closure runs with no


That is safe because the closure runs with no data-tree page locks held.

Are you sure about this statement?
As far as I know BPlusTree.invoke holds the write lock on leaf page while execution the closure, so it's not deadlock safe. Deadlock is possible between data-tree leaf pages since page can contain more than one entry (acquired entry lock is not enough protection).
Also, as far as I understand there can be deadlock on expiration: Expiration thread holds write lock on pending tree leaf page and reads data tree (with read lock). While addRow thread holds write lock on data tree leaf page and can concurrently evict data (remove rows) which remove ttl entries and requires write lock on pending tree leaf page.

Partially correct.
The premise — "BPlusTree.invoke holds the leaf write lock while running the closure" — is not supported by the code: BPlusTree.invokeDown releases the leaf read lock when read(...) returns and takes the leaf write lock only afterwards (in tryInsert/tryReplace/tryRemoveFromLeaf). Your broader point stands: entry tryLock covers only entry-level ordering and does not make eviction page-level lock-free.
The cross-tree (data→pending vs pending→data) residual risk with the TTL worker is real and documented in the code comment; a full fix is out of scope.
Please correct me if I'm wrong.

alex-plekhanov · 2026-09-10T15:38:07Z

     */
-    public void addRows(Collection<? extends CacheDataRow> rows,
-        IoStatisticsHolder statHolder) throws IgniteCheckedException {
+    public void addRows(Collection<? extends CacheDataRow> rows, IoStatisticsHolder statHolder) throws IgniteCheckedException {


addRows is only executed on rebalance. Why do we need this new code for rebalance but not for regular put?

reserving only the largest is insufficient

But you do exactly the same. Ensure free space for each row before any row is inserted, so only free space for largest row will be ensured. Other rows reservation will go through writeSinglePage.

Two points correct, one not.
(1) Rebalance-only: correct — regular puts use addRow (per-put reserve), both paths covered.
(2) Per-row loop == reserve for the largest: correct — a single reserve for the max row is equivalent.
(3) "Other rows go through writeSinglePage" — actually the batch path uses insertDataRows, whose per-page protection now comes from the shared takePageWithReserve re-reserve; the pre-reserve bounds the empty-pages counter up front.
Updated comment.

alex-plekhanov · 2026-09-10T15:57:59Z

+        }
+
        if (pageId == 0L) {
            pageId = allocateDataPage(row.partition());


There is also insertDataRows method where page can be requested on rows insert and oom can be fired.

Valid.
insertDataRows had the same allocateDataPage OOM-firing fallback without the lazy re-reserve, and since IgniteOutOfMemoryException is a RuntimeException, both methods mislabelled it as CorruptedFreeListException.
Made the batch trailing-fragment path symmetric via the shared takePageWithReserve, and both catch clauses rethrow OOM as-is so it is reported as OOM/critical failure, not corruption.

wernerdv changed the title ~~IGNITE-29032 Improvement of the pageEviction mechanism for in-memory~~ IGNITE-29032 Improvement of the page eviction mechanism for in-memory Sep 5, 2026

wernerdv requested a balanced review from Copilot September 5, 2026 16:09

Copilot started reviewing on behalf of wernerdv September 5, 2026 16:10 View session

Copilot AI reviewed Sep 5, 2026

View reviewed changes

wernerdv added 3 commits September 7, 2026 14:41

IGNITE-29032 Improvement of the pageEviction mechanism for in-memory

a6f98ea

address copilot comments

b159894

stabilize tests

22a93d4

wernerdv force-pushed the IGNITE-29032 branch 2 times, most recently from 9ffb5d4 to 956bfc5 Compare September 7, 2026 18:04

fix TOCTOU race

4d6ca55

wernerdv force-pushed the IGNITE-29032 branch from 956bfc5 to 4d6ca55 Compare September 7, 2026 19:28

wernerdv added 3 commits September 8, 2026 01:18

fix PDS tests

430c51a

fix broken tests

622b749

minor

ad83772

wernerdv force-pushed the IGNITE-29032 branch from 2ba28e5 to ad83772 Compare September 8, 2026 05:16

add benchmark

53a2c35

self-review

fb5b61e

wernerdv force-pushed the IGNITE-29032 branch from 9c8c1c1 to fb5b61e Compare September 9, 2026 05:00

alex-plekhanov reviewed Sep 10, 2026

View reviewed changes

wernerdv added 4 commits September 11, 2026 22:07

address review comments

965c9a8

Merge branch 'master' into IGNITE-29032

a8f9e51

improvement

db9de4d

minor

3e810dc

Uh oh!

Conversation

wernerdv commented Sep 4, 2026

The Contribution Checklist

Notes

Uh oh!

ignitetcbot commented Sep 4, 2026

TCBot Test Analysis

Possible Blockers (0)

New Tests (18)

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

🟡 Changes recommended

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

ignitetcbot commented Sep 8, 2026

TCBot Test Analysis

Possible Blockers (0)

New Tests (18)

Uh oh!

ignitetcbot commented Sep 9, 2026

TCBot Test Analysis

Possible Blockers (0)

New Tests (18)

Uh oh!

wernerdv commented Sep 9, 2026

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

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