Shenandoah GC myth or reality on the production ?

-

 Hi awesome community, 

 

Today I would like to share with you so small topic about usage Garbage Collectors in production (GC Strategies). 

At the moment, most popular in Hotspot JVM is G1 GC, after CMS  and rising star is Shenandoah GC.

 So in this topic I will describe how to migrate to Shenandoah GC. If you already on adoptJDK. 

And please, keep in your mind that small tutorial related to the Java 8, as my instance is not fully ready for new Java. (But please, be informed in your version Java works better and Shenandoah significantly improved as well.)

 

 Let’s start. 

  • All, my JDK. JRE used in the located in one place, like /usr/java and use frequently the symbolic links.

image.png

wget -c https://builds.shipilev.net/openjdk-shenandoah-jdk8/openjdk-shenandoah-jdk8-latest-linux-x86_64-release.tar.xz

  • Next step is extract into  
tar -xvf openjdk-shenandoah-jdk8-latest-linux-x86_64-release.tar.xz
  • Before change link let’s prepare our Jira instance. 
  1. Commented arguments, (about that arguments you can read here Flight Recorder)image.png

  2. JVM_SUPPORT_RECOMMENDED_ARGS="-XX:+UnlockCommercialFeatures -XX:+FlightRecorder"
  1. So let’s check my previous GC Strategy:
JVM_GC_STRATEGY="-server -Xss2048k -XX:+UseG1GC -XX:MaxGCPauseMillis=200 -XX:+UseStringDeduplication -XX:+PrintStringDeduplicationStatistics -XX:ConcGCThreads=8  -XX:ParallelGCThreads=24 -XX:ReservedCodeCacheSize=1024m -XX:+ExplicitGCInvokesConcurrent -XX:MaxMetaspaceSize=2048m"

2.1. Hence I removed the next parameters (it’s result of GC, thread, heap dump analysis tools analyzing): 

 -XX:+UseG1GC -XX:MaxGCPauseMillis=200 -XX:ConcGCThreads=8  -XX:ParallelGCThreads=24

2.2. Added

-XX:+UnlockExperimentalVMOptions -XX:+UseShenandoahGC

2.3. Finally, GC Strategy is: 

JVM_GC_STRATEGY="-server -Xss2048k -XX:+UnlockExperimentalVMOptions -XX:+UseShenandoahGC -XX:+UseStringDeduplication -XX:+PrintStringDeduplicationStatistics  -XX:MaxMetaspaceSize=2048m -XX:ReservedCodeCacheSize=1024m -XX:+ExplicitGCInvokesConcurrent "

 

About String Deduplication you can read here (Decrease Heap size)

  • Let’s run that easy pipeline command:
systemctl stop jira && unlink latest && ln -s j2sdk-image latest && systemctl start jira

As result in a few minutes you see in your logs that in need full info. (and Jira visible through browser)

 

That’s all.

Conclusion: Shenandoah has interesting extra options, but in my tests it was hard to understand a significant differences. But the default one, 

 

Hope it helps. 

If you interested I can share in the some metrics and charts of using Shenandoah and G1.

 

References:



Cheers,

Gonchik Tsymzhitov

Comments

Post a Comment

Popular posts from this blog

How only 2 parameters of PostgreSQL reduced anomaly of Jira Data Center nodes

-
Image
  Hello community, Gonchik Tsymzhitov is in touch. Today, I would like to share with you a short story about postgresqltuner.pl in usage.  So from previous articles I hope you configured the PostgreSQL server and postgresqltuner.pl.  After executing command, ./postgresqltuner.pl --ssd for ssd disks server. You can see like this output: postgresqltuner.pl version 1.0.1 [OK]      I can invoke executables Connecting to /var/run/postgresql:5432 database template1 as user 'postgres'... [OK]      The user account used by me for reporting has superuser rights on this PostgreSQL instance =====  OS information  ===== [INFO]    OS: linux Version: 5.4.17-2036.102.0.2.el8uek.x86_64 Arch: x86_64-linux-thread-multi [INFO]    OS total memory: 41.03 GB [OK]      vm.overcommit_memory is adequate: no memory overcommitment [INFO]    Running under a vmware hypervisor [INFO]    Currently used I/O sc...
Read more

Atlassian Community, let's collaborate and provide stats to vendors about our SQL index usage

-
  Hi!  Gonchik is in touch. Today I would like to get a little bit of help from you related SQL indexes on your Jira installation (Data Center, Server edition does not matter).    I would like to share a story about  PostgreSQL  and  Jira 8.13.3  ( JSD 4.13.3 )  release.  In that post I will be happy if you answer and comment about your instance indexes, as those info can help to improve the existing situation of heavy indexes.  Typically, it starts from improvement of configuration (e.g. using  https://postgresqltuner.pl/ ),  then query analyzing, adding indexes, after removing as heavy one (shortly  https://wiki.postgresql.org/wiki/Index_Maintenance ) :)    I use like this query in PostgreSQL: SELECT s.schemaname,        s.relname AS tablename,        s.indexrelname AS indexname,        pg_size_pretty(p...
Read more

Stories about detecting Atlassian Confluence bottlenecks with APM tool [part 1]

-
Image
  Hey! Gonchik, a lover of APM (application performance monitoring) tools, in particular Glowroot, is in touch. Today I will tell you how to find bottlenecks in Confluence On-Prem in the shortest possible time based on one industrial installation. We are faced with a situation where a large number of people simultaneously break into the knowledge base on Confluence On-Prem (during the certification), and the Confluence dies for some time. We immediately thought that the problem is precisely in the simultaneous number of visitors and we can immediately tweak the JVM, but it turned out that not everything is so simple. Below I will tell you how we found the real cause of the brakes and how we dealt with it. The main task: to conduct an audit and, on its basis, achieve performance improvements, especially in times of a large number of active users in the system. Of course, first of all, the hardware resources and OS configurations were checked, where no problems were identified. Howev...
Read more