[
https://issues.apache.org/jira/browse/IGNITE-23246?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17949969#comment-17949969
]
Denis Chudov edited comment on IGNITE-23246 at 5/7/25 11:41 AM:
----------------------------------------------------------------
We also had this case after some time of cluster running
{code:java}
Class Name
| Objects | Shallow Heap | Retained
Heap----------------------------------------------------------------------------------------------------------------------------------org.apache.ignite.internal.hlc.HybridTimestamp
| 154,174,254 | 3,700,182,096 | >=
3,700,185,528org.apache.ignite.internal.placementdriver.leases.Lease
| 76,995,000 | 3,079,800,000 | >= 6,775,206,072byte[]
| 431,933 |
2,416,352,296 | >= 2,416,352,296java.util.HashMap$TreeNode
| 36,002,191 | 2,016,122,696 | >=
6,912,296,824java.util.concurrent.CompletableFuture
| 77,576,020 | 1,861,824,480 | >=
14,268,283,984org.apache.ignite.internal.placementdriver.negotiation.LeaseAgreement
| 76,983,201 | 1,847,596,824 | >=
10,469,636,984java.util.HashMap$Node
| 41,137,551 | 1,316,401,632 | >=
6,918,308,816java.util.HashMap$Node[]
| 46,888 | 430,727,448 | >=
14,261,786,344java.lang.Object[]
| 219,348 | 44,844,248 | >=
147,034,680org.apache.ignite.raft.jraft.core.FSMCallerImpl$ApplyTask
| 557,056 | 35,651,584 | >= 35,651,912long[]
| 107,889 |
27,226,024 | >=
27,226,024org.apache.ignite.raft.jraft.core.ReadOnlyServiceImpl$ReadIndexEvent
| 524,288 | 25,165,824 | >=
25,165,824org.apache.ignite.raft.jraft.core.NodeImpl$LogEntryAndClosure
| 524,288 | 25,165,824 | >= 25,165,824int[]
| 24,553 |
16,720,264 | >=
16,720,264java.util.concurrent.locks.ReentrantReadWriteLock$NonfairSync
| 335,599 | 16,108,752 | >=
21,650,144java.util.concurrent.CompletableFuture$UniWhenComplete
| 362,137 | 14,485,480 | >=
14,257,734,296java.util.concurrent.locks.ReentrantReadWriteLock
| 329,668 | 7,912,032 | >=
38,427,896java.util.concurrent.ConcurrentHashMap
| 118,581 | 7,589,184 | >=
56,428,576java.lang.ThreadLocal$ThreadLocalMap$Entry
| 222,871 | 7,131,872 | >=
9,195,216org.apache.ignite.internal.table.distributed.gc.MvGc$$Lambda+0x00007f4730fb5808|
277,300 | 6,655,200 | >=
6,655,200java.util.concurrent.ScheduledThreadPoolExecutor$ScheduledFutureTask
| 90,777 | 6,535,944 | >= 15,752,031,576Total: 21 of 21,880
entries; 21,859 more | 474,012,399 |
17,120,715,168 | {code}
was (Author: denis chudov):
We also had this case after some time of cluster running
{code:java}
Class Name
| Objects | Shallow Heap | Retained
Heap----------------------------------------------------------------------------------------------------------------------------------org.apache.ignite.internal.hlc.HybridTimestamp
| 154,174,254 | 3,700,182,096 | >=
3,700,185,528org.apache.ignite.internal.placementdriver.leases.Lease
| 76,995,000 | 3,079,800,000 | >= 6,775,206,072byte[]
| 431,933 |
2,416,352,296 | >= 2,416,352,296java.util.HashMap$TreeNode
| 36,002,191 | 2,016,122,696 | >=
6,912,296,824java.util.concurrent.CompletableFuture
| 77,576,020 | 1,861,824,480 | >=
14,268,283,984org.apache.ignite.internal.placementdriver.negotiation.LeaseAgreement
| 76,983,201 | 1,847,596,824 | >=
10,469,636,984java.util.HashMap$Node
| 41,137,551 | 1,316,401,632 | >=
6,918,308,816java.util.HashMap$Node[]
| 46,888 | 430,727,448 | >=
14,261,786,344java.lang.Object[]
| 219,348 | 44,844,248 | >=
147,034,680org.apache.ignite.raft.jraft.core.FSMCallerImpl$ApplyTask
| 557,056 | 35,651,584 | >= 35,651,912long[]
| 107,889 |
27,226,024 | >=
27,226,024org.apache.ignite.raft.jraft.core.ReadOnlyServiceImpl$ReadIndexEvent
| 524,288 | 25,165,824 | >=
25,165,824org.apache.ignite.raft.jraft.core.NodeImpl$LogEntryAndClosure
| 524,288 | 25,165,824 | >= 25,165,824int[]
| 24,553 |
16,720,264 | >=
16,720,264java.util.concurrent.locks.ReentrantReadWriteLock$NonfairSync
| 335,599 | 16,108,752 | >=
21,650,144java.util.concurrent.CompletableFuture$UniWhenComplete
| 362,137 | 14,485,480 | >=
14,257,734,296java.util.concurrent.locks.ReentrantReadWriteLock
| 329,668 | 7,912,032 | >=
38,427,896java.util.concurrent.ConcurrentHashMap
| 118,581 | 7,589,184 | >=
56,428,576java.lang.ThreadLocal$ThreadLocalMap$Entry
| 222,871 | 7,131,872 | >=
9,195,216org.apache.ignite.internal.table.distributed.gc.MvGc$$Lambda+0x00007f4730fb5808|
277,300 | 6,655,200 | >=
6,655,200java.util.concurrent.ScheduledThreadPoolExecutor$ScheduledFutureTask
| 90,777 | 6,535,944 | >= 15,752,031,576Total: 21 of 21,880
entries; 21,859 more | 474,012,399 |
17,120,715,168 | {code}
> Lease agreements might leak
> ---------------------------
>
> Key: IGNITE-23246
> URL: https://issues.apache.org/jira/browse/IGNITE-23246
> Project: Ignite
> Issue Type: Bug
> Reporter: Vladislav Pyatkov
> Assignee: Denis Chudov
> Priority: Major
> Labels: ignite-3
>
> h3. Motivation
> Agreements are used to consider a lease, but unlike leases, the agreements
> are stored in memory.
> {code}
> /** Lease agreements which are in progress of negotiation. */
> private final Map<ReplicationGroupId, LeaseAgreement> leaseToNegotiate;
> {code}
> These agreements exist only in PD active-actor, but in the case where the
> active-actor does not change for a long time, we would get agreements for
> removed replication groups.
> The type of agreement is useless but consumes a lot of memory.
> h3. Definition of done
> Agreemens should be deleted ether with replication group stop or when the
> agreement is expired.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
