-
-
Save huynhbaoan/1bb4cd87313f56bc63cc4e19ce5c3ec7 to your computer and use it in GitHub Desktop.
| Recently, I use JMXterm to collect info about Java JVM via JMX and MBeans. Here is a short note. | |
| - Download JMXterm | |
| https://sourceforge.net/projects/cyclops-group/files/jmxterm/1.0.0/jmxterm-1.0.0-uber.jar/download | |
| - Run JMXterm | |
| java -jar jmxterm-1.0.0-uber.jar --url localhost:<jmx listen port> | |
| - All MBeans from java.lang | |
| $>domain java.lang | |
| $>beans | |
| java.lang:name=CMS Old Gen,type=MemoryPool | |
| java.lang:name=Code Cache,type=MemoryPool | |
| java.lang:name=CodeCacheManager,type=MemoryManager | |
| java.lang:name=Compressed Class Space,type=MemoryPool | |
| java.lang:name=ConcurrentMarkSweep,type=GarbageCollector | |
| java.lang:name=Metaspace Manager,type=MemoryManager | |
| java.lang:name=Metaspace,type=MemoryPool | |
| java.lang:name=Par Eden Space,type=MemoryPool | |
| java.lang:name=Par Survivor Space,type=MemoryPool | |
| java.lang:name=ParNew,type=GarbageCollector | |
| java.lang:type=ClassLoading | |
| java.lang:type=Compilation | |
| java.lang:type=Memory | |
| java.lang:type=OperatingSystem | |
| java.lang:type=Runtime | |
| java.lang:type=Threading | |
| - Usually, we want to know how many time a garbage collection happen and how long a garbage collection is, so we use 2 MBeans | |
| java.lang:name=ParNew,type=GarbageCollector (young GC) | |
| java.lang:name=ConcurrentMarkSweep,type=GarbageCollector (in this case, I use CMS as major GC) | |
| Now, let's explore inside | |
| $>bean java.lang:name=ParNew,type=GarbageCollector | |
| If Name has space, follow it by \\ | |
| $>bean java.lang:name=PS\\ Scavenge,type=GarbageCollector | |
| $>info | |
| #mbean = java.lang:name=ParNew,type=GarbageCollector | |
| #class name = sun.management.GarbageCollectorImpl | |
| # attributes | |
| %0 - CollectionCount (long, r) | |
| %1 - CollectionTime (long, r) | |
| %2 - LastGcInfo (javax.management.openmbean.CompositeData, r) | |
| %3 - MemoryPoolNames ([Ljava.lang.String;, r) | |
| %4 - Name (java.lang.String, r) | |
| %5 - ObjectName (javax.management.ObjectName, r) | |
| %6 - Valid (boolean, r) | |
| #there's no operations | |
| # notifications | |
| %0 - javax.management.Notification(com.sun.management.gc.notification) | |
| Here, we care about CollectionCount and CollectionTime. Let's do a simple query. | |
| $>get CollectionCount | |
| #mbean = java.lang:name=ParNew,type=GarbageCollector: | |
| CollectionCount = 546; | |
| So, we can do something similar to fetch the query result to a time-series database, then have grafana draw a graph for us | |
| - Query by Zabbix Zabbix jmxagent | |
| Define an item with a key like this: | |
| jmx["java.lang:type=GarbageCollector,name=ParNew","CollectionCount"] | |
| - Query by collectd JMX plugin. | |
| Well, this is more complicated to define, because of the plugin configuration. We can take an example here: https://gist.github.com/huynhbaoan/e2cf40654649237cd6ca495ce5a9e13b | |
| And the offical document: https://collectd.org/wiki/index.php/Plugin:GenericJMX | |
| I will update collectd configurations later, if we decide to use this for our monitor solution. | |
| - TL; DR: | |
| Here is a short list of key to apply on Zabbix jmx agent. You can use these MBeans info to configure collectd plugin. | |
| jmx["java.lang:type=GarbageCollector,name=ParNew","CollectionCount"] | |
| jmx["java.lang:type=GarbageCollector,name=ParNew","CollectionTime"] | |
| jmx["java.lang:type=GarbageCollector,name=ConcurrentMarkSweep","CollectionCount"] | |
| jmx["java.lang:type=GarbageCollector,name=ConcurrentMarkSweep","CollectionTime"] | |
| jmx["java.lang:type=Memory","HeapMemoryUsage.used"] | |
| jmx["java.lang:type=Memory","HeapMemoryUsage.max"] | |
| jmx["java.lang:type=Memory","NonHeapMemoryUsage.used"] | |
| jmx["java.lang:type=Memory","NonHeapMemoryUsage.max"] | |
| jmx["java.lang:type=MemoryPool,name=Par Eden Space","Usage.used"] | |
| jmx["java.lang:type=MemoryPool,name=Par Survivor Space","Usage.used"] | |
| jmx["java.lang:type=MemoryPool,name=CMS Old Gen","Usage.used"] | |
| jmx["java.lang:type=MemoryPool,name=CMS Perm Gen","Usage.used"] | |
| jmx["java.lang:type=MemoryPool,name=Code Cache","Usage.used"] | |
| jmx["java.lang:type=Threading","DaemonThreadCount"] | |
| jmx["java.lang:type=ClassLoading","LoadedClassCount"] | |
| jmx["java.lang:type=ClassLoading","UnloadedClassCount"] | |
| jmx["java.lang:type=OperatingSystem","ProcessCpuLoad"] | |
| jmx["java.lang:type=Threading","ThreadCount"] | |
| jmx["java.lang:type=Compilation","TotalCompilationTime"] |
http://www.fasterj.com/articles/oraclecollectors1.shtml
Young generation collectors
Copy (enabled with -XX:+UseSerialGC)
the serial copy collector, uses one thread to copy surviving objects from Eden to Survivor spaces and between Survivor spaces until it decides they've been there long enough, at which point it copies them into the old generation.
PS Scavenge (enabled with -XX:+UseParallelGC)
the parallel scavenge collector, like the Copy collector, but uses multiple threads in parallel and has some knowledge of how the old generation is collected (essentially written to work with the serial and PS old gen collectors).
ParNew (enabled with -XX:+UseParNewGC)
the parallel copy collector, like the Copy collector, but uses multiple threads in parallel and has an internal 'callback' that allows an old generation collector to operate on the objects it collects (really written to work with the concurrent collector).
G1 Young Generation (enabled with -XX:+UseG1GC)
the garbage first collector, uses the 'Garbage First' algorithm which splits up the heap into lots of smaller spaces, but these are still separated into Eden and Survivor spaces in the young generation for G1.
Old generation collectors
MarkSweepCompact (enabled with -XX:+UseSerialGC)
the serial mark-sweep collector, the daddy of them all, uses a serial (one thread) full mark-sweep garbage collection algorithm, with optional compaction.
PS MarkSweep (enabled with -XX:+UseParallelOldGC)
the parallel scavenge mark-sweep collector, parallelised version (i.e. uses multiple threads) of the MarkSweepCompact.
ConcurrentMarkSweep (enabled with -XX:+UseConcMarkSweepGC)
the concurrent collector, a garbage collection algorithm that attempts to do most of the garbage collection work in the background without stopping application threads while it works (there are still phases where it has to stop application threads, but these phases are attempted to be kept to a minimum). Note if the concurrent collector fails to keep up with the garbage, it fails over to the serial MarkSweepCompact collector for (just) the next GC.
G1 Mixed Generation (enabled with -XX:+UseG1GC)
the garbage first collector, uses the 'Garbage First' algorithm which splits up the heap into lots of smaller spaces.
JVM heap
Source: https://www.linkedin.com/pulse/java-virtual-machine-changes-78-9-kunal-saxena/