Hi All, Thanks for your reply. The committed memory in my scenario increases from 1GB to 12GB but does not decrease.
I have restarted grid 5 days ago and Java heap usage trends are :- [15:20:55,784] Heap [used=7659MB, free=51.57%, comm=12192MB] [15:19:55,776] Heap [used=5763MB, free=63.56%, comm=12192MB] [15:18:55,768] Heap [used=10505MB, free=33.58%, comm=12192MB] [15:17:55,762] Heap [used=4761MB, free=69.89%, comm=12192MB] Thanks & Regards, Charlin On Tue, 17 Feb 2026 at 14:33, Stephen Darlington <[email protected]> wrote: > True, but that's the case with pretty much all software. > > On Sat, 14 Feb 2026 at 20:28, Humphrey <[email protected]> wrote: > >> Ignite also doesn’t release the off heap memory it allocated. >> >> On 12 Feb 2026, at 12:52, Stephen Darlington <[email protected]> >> wrote: >> >> >> We should probably improve the documentation around this issue, as it's a >> common misunderstanding. >> >> Simplifying slightly, Ignite has two kinds of memory: on-heap and >> off-heap. The memory you configured in XML is the off-heap memory. That's >> where Ignite stores your data. There's also the Java heap, which you can >> think of as its working set. I don't know the default in your environment. >> >> You need to add the types together (and then add a bit) to find Ignite's >> total memory footprint. >> >> See here for how to configure the heap memory: >> https://ignite.apache.org/docs/ignite2/latest/perf-and-troubleshooting/memory-tuning >> >> On Thu, 12 Feb 2026 at 10:27, Charlin S <[email protected]> wrote: >> >>> Hi All, >>> >>> I've encountered an issue where memory usage is spiking on the Ignite >>> server, and the memory is not being released back to the OS. I am hosting >>> three grids on this server: two use replication cache mode with no expiry >>> policy, and the third uses partition cache mode with an expiry policy. >>> >>> This issue first appeared with Amazon Linux 2023.9.20251027 and `openjdk >>> version "1.8.0_402-402"`. We subsequently upgraded to Java 17, focusing on >>> the primary JVM options designed to encourage the release of unused heap >>> memory to the operating system. We start Ignite using `ignite.sh >>> config/grid1.xml`. >>> Please suggest ways to finetune the configuration for better memory >>> management. >>> >>> The current upgraded version details are: >>> >>> ``` >>> openjdk version "17.0.17" 2025-10-21 LTS >>> OpenJDK Runtime Environment Microsoft-12574440 (build 17.0.17+10-LTS) >>> OpenJDK 64-Bit Server VM Microsoft-12574440 (build 17.0.17+10-LTS, mixed >>> mode, sharing) >>> >>> >>> ``` >>> >>> Observations: >>> >>> * One Ignite grid (partition mode, 2-cluster) uses 19GB of memory, but >>> its data region is only 18GB. >>> * Another grid (replication mode, 2-cluster) uses 21GB of memory, but >>> its data region is only 10GB. >>> >>> >>> >>> bean xml file >>> <?xml version="1.0" encoding="UTF-8"?> >>> <beans xmlns="http://www.springframework.org/schema/beans"; >>> xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance"; >>> xmlns:util="http://www.springframework.org/schema/util"; >>> xsi:schemaLocation="http://www.springframework.org/schema/beans >>> >>> http://www.springframework.org/schema/beans/spring-beans.xsd >>> http://www.springframework.org/schema/util >>> >>> http://www.springframework.org/schema/util/spring-util.xsd";> >>> <bean class="org.apache.ignite.configuration.IgniteConfiguration"> >>> <property name="clientMode" value="false"/> >>> <property name="clientFailureDetectionTimeout" value="120000"/> >>> <property name="igniteInstanceName" value="Productiongrid"/> >>> <property name="longQueryWarningTimeout" value="4000"/> >>> <property name="failureDetectionTimeout" value="120000"/> >>> <property name="segmentationPolicy" value="RESTART_JVM"/> >>> <property name="segmentCheckFrequency" value="20000"/> >>> <property name="segmentationResolveAttempts" value="10"/> >>> <property name="discoverySpi"> >>> <bean class="org.apache.ignite.spi.discovery.tcp.TcpDiscoverySpi"> >>> <property name="localPort" value="55500"/> >>> <property name="ipFinder"> >>> <bean >>> class="org.apache.ignite.spi.discovery.tcp.ipfinder.vm.TcpDiscoveryVmIpFinder"> >>> <property name="addresses"> >>> <list> >>> <value>IP:55500</value> >>> <value>IP:55500</value> >>> </list> >>> </property> >>> </bean> >>> </property> >>> </bean> >>> </property> >>> <property name="dataStorageConfiguration"> >>> <bean >>> class="org.apache.ignite.configuration.DataStorageConfiguration"> >>> <property name="defaultDataRegionConfiguration"> >>> <bean >>> class="org.apache.ignite.configuration.DataRegionConfiguration"> >>> <property name="name" >>> value="Dynamic_Data_Region"/> >>> <property name="initialSize" value="#{1L * 1024 >>> * 1024 * 1024}"/> >>> <property name="maxSize" value="#{18L * 1024 * >>> 1024 * 1024}"/> >>> <property name="pageEvictionMode" >>> value="RANDOM_2_LRU"/> >>> <property name="evictionThreshold" value="0.7"/> >>> <property name="emptyPagesPoolSize" >>> value="65536"/> >>> </bean> >>> </property> >>> </bean> >>> </property> >>> <property name="communicationSpi"> >>> <bean >>> class="org.apache.ignite.spi.communication.tcp.TcpCommunicationSpi"> >>> <property name="sharedMemoryPort" value="-1"/> >>> <property name="localPort" value="55605"/> >>> </bean> >>> </property> >>> </bean> >>> </beans> >>> >>> Thanks & Regards, >>> Charlin >>> >>>
