Feature added autotune

[alexbacchin]

I am using this cookbook for the last few days and setting up jvm memory can be tricky if you don't have much Java experience.
As this was based on the Jira cookbook, I have ported the autotune recipe to the confluence cookbook.
Even the default setting for autotune is false. When used, it will make confluence run better and utilize better the server memory.

[legal90]

Hi @alexbacchin
Thank you for the contribution! Yes, setting up the memory for java applications is tricky.
But currently, I'm not sure that we need to add such autotune logic to this community cookbook.

My concerns:

• There are no any rules or "silver bullet" examples of JRE memory settings for Confluence. Atlassian recommends tuning Xms & Xmx according to the server load, not to the total memory available on the node. Actually, these settings are really specific for each concrete server instance in production.
• In Confluence 6 there was added a new component, Synchrony (for Collaborative Editing feature), which runs as a separate java container. Its Xmx is set to 1g, so it means that there are no any simple formulae applicable for everyone.

I really appreciate your effort and honestly, I would be happy if we can use such universal formulae everywhere. But unfortunately, in a real production we need to tune such settings manually for each concrete installation.

@bflad @patcon, what do you think?

[alexbacchin]

Hi @legal90,
I believe the idea with the autotune is to provide an good enough JVM setting so the admin does not have to worry about this initially. It was not designed to be the perfect setting, but it can be a starting point.
Also if you consider scalability, as you Confluence environment grows memory is added to the host (or VM) but JMV settings can be changed automatically by Chef with the autotune suggestions.
As it is disabled by default, we are not enforcing this to be used.

In regards the Synchrony, I can change the calculation to leave memory for it.

[patcon]

Seems reasonable in principle :) although have not tested, and not sure the maintenance burden

[anuriq]

Overall, this is a lot of code for something which could be set by admin after 1 min thinking. Do we really need this?

[alexbacchin]

I performed the updates based on @anuriq feedback.
In regards the usefulness of this feature. I have found this quite useful when testing different server memory settings. Therefore, the JVM memory adapts when you need to vertically scale the server.
Nevertheless, if you all still don't feel useful, I will keep it on my fork. No worries at all.

[anuriq]

thanks!

[mvdkleijn]

Cool to see this go in. I authored the auto tune feature in the Jira cookbook. For future readers: it was heavily based on the auto tune feature of the Postgresql cookbook. 😄

[lock]

This thread has been automatically locked since there has not been any recent activity after it was closed. Please open a new issue for related bugs.