#Poor server performance (Spark results question)

17 messages · Page 1 of 1 (latest)

carmine cape
#

When there are 5+ people on my server, the performance starts getting bad. I ran Spark and it shows that the process max memory is 3.5GB. How can this be? How do I increase this? My server has 64GB memory available.

#

Changing the launcher parameters doesn't seem to help or do much

#

Java process taking up 4.5GB

reef juniper
carmine cape
#

Hmm. Am I not reading it correctly? I have it set at 8GB max, why does the max show as 3.5GB in Spark?

#

What is recommended? Can I just put the min and max at 12GB? Or more?

reef juniper
#

min is the ram it allocates when it starts
when it runs out of ram and has to garbage collect, it tries to clear from the used ram. if it cant clear below a threshold, it allocates more ram.

max is the maximum ram that it is allowed to allocate for heap. so min is the starting, max is the "dont go higher than this for Heap allocations"

#

if you set both to the same value, then it allocates the value (min) and does not try to allcate more (max)

carmine cape
#

I see. Are there any recommendations when it comes to other settings?
I'll just start with 12GB and see how it goes for the memory.

reef juniper
carmine cape
#

Awesome, will give it a look. Thanks for the quick responses and explanations!

carmine cape
#

Above 5 people we still experience issues. But dont know if anything here can be improved :/ Mobticks seem very expensive(?)
The memory issues are fixed though!

carmine cape
#

Unfortunately i haven't managed to make any progress regarding this.
If anyone has any idea what to look at or things i could try still, it's much appreciated.

vocal turtle
#

Of Lingaurdium’s suggestions, which have you tried?

carmine cape
carmine cape
#

Adoptium without any GC seems to be an improvement when it comes to TPS, but we're not there yet.
Still, mobs and other entities seem to use more resources than they should

vocal turtle
#

Just checking... what Craftoria version is this? That lambda function taking up a large part of the time is vaguely reminding me of an older issue.