Eden space increasing by 10Mb every minute and GC'ing back t

[khooke]

I'm running JBoss 4.0.1sp1 on Fedora Core4, with JDK 1.5.0_05.

I have a low traffic web app deployed to the server that gets approx 6 page hits an hour. The JVM is configured with -Xms128m -Xmx128m.

Using JConsole I see the eden VM space increase by 10Mb each minute and then get collected back down to near 0. This is a regular pattern each minute. The survivor space is next to 0 and tenured gen is also constant at around 34Mb.

Even during periods of no activity on my server, I am stil seeing this contunuous saw-tooth pattern of memory use in the eden space - it goes from 0 to 10Mb in 1 min, and then is collected back down to 0 again.

Is this normal behavior?

[markgowdy]

Hi

I haven't checked this on JBoss but this pattern is consistent with the heap usage patterns on other containers and application servers. It is due to the application server threads running in the background and the objects allocated/GC'd for that purpose. This is not something to be concerned about, unless the amount of heap in use increases over time (ie, look at the heap usage for each valley before/after GC activity and if this increases, then this could point to an application server/application leak). Also, 0-10Mb is not a particularly large delta so I would not be too concerned here. If it was going from 0-100+Mb then I would perform additional analysis.

Since this is JBoss, you may wish to look at the AppServer slimming/tuning wiki to see if there are any services you can turn off/remove. This will have an impact on performance, and probably also on the heap usage.

Hope this helps.

Mark

[genman]

One thing that's run periodically is a URL deployment scanner. This may itself generate 10MB of garbage every 10 seconds or so.