Empty Jira Software installation eats 1+ GB memory?

[Norbert Virth]

Hi,

I've just installed Jira (v7.5.2). It's empty, no project in it yet; I'm the only user.
I was shocked when I saw it has already eaten 1+ GB memory.

I tried, with the same results:

• On Windows 10, with an msi installation
• On Ubuntu, with docker*
• On CentOS, with docker*

*I've tried both of these docker imagges:
cptactionhank/atlassian-jira
blacklabelops/jira

I can read in the official installation requirements that Jira should work with 2GB RAM up to 100 projects, 1000 issues and 100 users; but actually it's hard to believe.
For example, in Ubuntu, my empty Jira starts with 1.3GB. I only clicked some in the admin section, did not add any user or project or issue; it went up to 1.5GB!
Actually, the Jira's Java process was started with "-Xmx1g", but it seems to me the process somehow ignores it. I tried to decrease it to "-Xmx512m", did not seem to change anything.

Am I doing smg wrong, or is this normal?
Can one somehow decrease the memory footprint for small Jira instances?

Thank you!

[Andy Heinzer · Accepted solution]

Hi Norbert,

I understand that you are installing Jira and have some concerns over the memory usage you see when compared against that documentation.  Thanks for bringing this to our attention.

I would say that the document we have for that is actually out of date.   As such I have created an internal request to update that page to more accurately reflect the current server production install requirements.

In the mean time I would instead recommend checking out the Jira Sizing Guide.  This page provides a better current expectation of the hardware you will need to run Jira, and do so smoothly.

I personally am not very surprised that a base Jira installation is going to use 1GB of system memory from the start.   I understand if you were looking to lower the amount of memory Jira using.  This might be possible to do in a limited fashion, however when you set this value to be too low, Jira's performance tends to drop badly.  This is because Jira is running inside of a Java environment.  As such it is still bound to the garbage collection process to reclaim memory that all Java applications have to use.   If Java is having to perform frequently full gc events, Jira becomes unresponsive while these are happening.   I would recommend Confluence's Performance Tuning Guide.  The steps and ideas there can still largely be applied to tuning Jira as well.

You mentioned trying to the change xmx value but Jira did not seem to accept this.   Increasing memory in Jira has steps you can follow in order to set the Xmx and Xms values.  But if you're running in a Windows environment, the steps to do this are slightly different depending on if you're starting Jira as a service vs starting Jira via the .bat files.   Either way, after you make this change, you always have to restart Jira for these changes to take effect. 

In general, I wouldn't recommend giving Jira less than the default memory max (Xmx).  There are other performance problems that can happen when providing the application with not enough memory.

Regards,
Andy