Hello

Not sure if the cause is that I used to have the swap plugin, but now I have this issue. kcompactd0 kicks in and uses a lot of CPU very often killing VM performance despite CPU pinning.

 

I also noticed that I have bery little free ram?

 

MiB Mem :  63665.3 total,    862.2 free,  10738.7 used,  52064.4 buff/cache
MiB Swap:      0.0 total,      0.0 free,      0.0 used.  35281.2 avail Mem

 

Is this normal behavior?

s4-unraid2-diagnostics-20240127-2055.zip

9 hours ago, fr500 said:

I also noticed that I have bery little free ram?

This is the important part:

9 hours ago, fr500 said:

35281.2 avail Mem

 

alright, so is kswapd even available on unraid by default? or did I install something and screw myself up?

Edited January 29, 20242 yr by fr500

Did you reboot after uninstalling the plugin?

Yeah..

I'm thinking the plugin may alter some sysctls, because I was able to stop this behavior by setting these flags:

 

sysctl -w vm.swappiness=0
sysctl -w vm.compaction_proactiveness=0
sysctl -w vm.extfrag_threshold=1000

 

Swappiness only didn't stop it, but after the other two it did

2 minutes ago, fr500 said:

Yeah..

I'm thinking the plugin may alter some sysctls, because I was able to stop this behavior by setting these flags:

 

sysctl -w vm.swappiness=0
sysctl -w vm.compaction_proactiveness=0
sysctl -w vm.extfrag_threshold=1000

 

Swappiness only didn't stop it, but after the other two it did

 

So I guess I gotta check what the default values were for these

After a reboot everything should go back to defaults, default for this one for example is 60

5 minutes ago, fr500 said:

sysctl -w vm.swappiness=0

 

Yeah that's the case for me too...

So weird then, I don't know why kcompactd0 is pegging one core every few minutes when I have so much ram free

This is so annoying, the only way I have managed to avoid kcompactd0 from kicking in and messing things up is

 

at 01:00 a, I do

 

echo never > /sys/kernel/mm/transparent_hugepage/defrag
sysctl -w vm.compaction_proactiveness=0
sysctl -w vm.extfrag_threshold=1000

 

Then at 00:00 I do

 

sysctl -w vm.compaction_proactiveness=20
sysctl -w vm.extfrag_threshold=500
echo always > /sys/kernel/mm/transparent_hugepage/defrag
sysctl -w vm.compact_memory=1

 

Now instead of kcompactd0 I get khugepaged going overboard for a while every 30 minutes or so but it's definitely not as annoying as kcompactd0 hoarding one CPU for 10 minutes

 

Guess I'll need to start over with my seup, I probably installed something the screwed something up