Infinity 1.8.2 RAM nomming.

[Shevron]

Short history:

- Was running Infinity 1.4.
- Upgraded to 1.5 which crashed server every 20 minutes, so rolled back to 1.4 and let it be for a few months.

I have yesterday upgraded my server and map to 1.8.2 (recommended ), and it's stable. Happy times.

However I'm noticing this version is eating a lot more RAM than 1.4 used to. I run MC on a FragNet server with 3GB RAM (with Fragnet you pay according to RAM allocated), and while with 1.4 sat happily at 80% memory usage with 5 players online, 1.8 is eating up a 100% of it, and nobody is even logged in.

Anyone noticed this?

Mystcraft is done with its profiling, as I suspected it to be the culprit, but it isn't.

 

[Harvest88]

There are multiple reasons why an update could increase your memory utilization considerably. For instance updating from 1.4 to 1.8.2 there ought to be additional mods added to the pack. The more mods there is, the more RAM it will need on "idle".

 

[tofuturkey]

I'm not running a server, just singleplayer, so there are probably differences there. However, when playing singleplayer with 8GB allocated, the game only uses approximately 2.5GB. Also updated to 1.8.2... While there are definitely new mods added since 1.4, I doubt that the server would be using a whole extra .5GB just because it is a server and not singleplayer. Probably something else using up the memory.

 

[Cptqrk]

How big is your server Shevron? Are there a lot of big builds? What's your TPS running at?

Could be one mod that got an update that has a glitch..

 

[Player]

Install my sampler mod from the Fastcraft thread and run "/sampler memory --gc". I doubt your readings actually represent the real memoy usage.

 

[Harvest88]

Install my sampler mod from the Fastcraft thread and run "/sampler memory --gc". I doubt your readings actually represent the real memoy usage.

Actually McMyAdmin is pretty good when I last used it. It's Multicraft that you may be thinking of and yes that panel's metering is honestly not the best..

 

[Player]

Try and compare it. The way the JVM manages its heap memory makes it difficult to separate the observable memory footprint into what's actually in use and what can be freed at any time. The latter part and therefore also the overall value have little informative value as heuristics and configuration influence it too much.