High CPU load with smb

[dehein2]

Hi all,

 

i have a setup with a cache disk (enrypted) and an array (encrypted). When i copy over files to the new server via smb (just using the windows explorer) i have very high CPU loads and the system becomes unresponsive.

Could it be the enryption? I have a i5-6600K CPU, so AES NI should be available.

 

Best

 Dennis

[JorgeB]

If you have dockers running stop the docker engine and see if there's any difference, if yes get a dedicated device for appdata.

[dehein2]

- The dockers don't have a high CPU load

 --> but when i completely disable docker the CPU load is still high (maybe slightly lower)

- i already have an extra pool for appdata

 --> appdata - Prefer : Data

 --> Multimedia - Yes : Cache

 --> the array is the same

 

 

Edited May 12, 2021 by dehein2

[dehein2]

i just checked sth. which looks good

 

grep -o aes /proc/cpuinfo
aes
aes
aes
aes


sort -u /proc/crypto | grep module
module       : aesni_intel

 

still high load

Edited May 13, 2021 by dehein2

[dehein2]

just 2  pictures:

 

 

not sure if i interpret top correctly, but the load seems to be lower there. But as the system still is unresponsive i don't think that's right

Edited May 13, 2021 by dehein2

[JorgeB]

Top doesn't show i/o load, dashboard does.

[dehein2]

24 minutes ago, JorgeB said:

Top doesn't show i/o load, dashboard does.

 I’m not sure what that means

 

does it mean that it’s normal that in top the load is lower?

 

what does that mean for my problem?

Edited May 13, 2021 by dehein2

[JorgeB]

1 minute ago, dehein2 said:

does it mean that it’s normal that in top the load is lower?

Yes.

 

1 minute ago, dehein2 said:

what does that mean for my problem?

Not much.

[dehein2]

Hehe ok.. than I’m open for help ;)))

[JorgeB]

If it's the same with the docker stopped (not the dashboard load, that is normal to be high, if the server turns unresponsive), don't have many more ideas, you could try without encryption just to test, but it shouldn't be an issue with your CPU.

[dehein2]

ok, so 100% load in the dashboard is "normal"

[JorgeB]

I wouldn't say 100% is normal, but high dashboard load during a transfer is normal, because of the i/o load, but unless Unraid/GUI gets unresponsive during that it's nothing to worry about.

[dehein2]

maybe 100% was a bit over-exaggerated

I'll observe it a bit. Currently i would say its high but not unresponsive. Maybe it was just a temporary problem

 

Thanks

[dehein2]

one more observation:

if i copy directly to the array (cahe is full ;)) the CPU load in the dashboard is a lot lower... both the array and the cache are enrypted. The array conistst out of 2 17TB HDDs and the cache of 2 SSDs....

[JorgeB]

What model SSDs? Low end DRAMless TLC SSDs can cause higher i/o load.

[dehein2]

I have  2 Samsung 870 QVO 1tb

[JorgeB]

Those are QLC, after the small SLC cache is used they are terribly slow at writing, live with it or replace them with the much faster EVO.

[dehein2]

Ok, I have 2 Samsung SSD 870 EVO I could use. 
 

can I just replace the cache drives disks? Or do I empty it with the mover and delete it and create a new one with the new discs and change it in the share settings?

[JorgeB]

Cache replacement is broken in v6.9.x, you'd need to do it manually or that:

10 minutes ago, dehein2 said:

Or do I empty it with the mover and delete it and create a new one with the new discs and change it in the share settings?

 

[dehein2]

whats the safest way to do it -> set the share to cache = no, start the mover and wait?

[JorgeB]

Cache=yes for moving to array:

https://forums.unraid.net/topic/46802-faq-for-unraid-v6/?do=findComment&comment=537383

 

Note that docker/VM service need to be disable, it won't move open files.

[dehein2]

Ok thanks 

 

[dehein2]

On 5/13/2021 at 7:03 PM, JorgeB said:

Cache=yes for moving to array:

https://forums.unraid.net/topic/46802-faq-for-unraid-v6/?do=findComment&comment=537383

 

Note that docker/VM service need to be disable, it won't move open files.

Hi,

 

maybe i dud sth. wrong. I created the new pool. Set all shars with old pool to "yes". Started the mover.

But there are still some files on the cache (a few i could just ignore and a docker.img). Docker/VM were disabled.

[JorgeB]

Enable mover logging, run the mover, post diags.

[dehein2]

for the system folder with the docker.imh. i changed the system share to another cache earlier - maybe the files were just abandoned?

for the files from the multimedia share:

May 17 19:39:51 deheinServer root: mover: started
May 17 19:39:51 deheinServer move: move: file /mnt/cache/Multimedia/Fitness/Intensiv Yoga/3 - Programm B.mkv
May 17 19:39:51 deheinServer move: move_object: /mnt/cache/Multimedia/Fitness/Intensiv Yoga/3 - Programm B.mkv File exists
May 17 19:39:51 deheinServer move: move: file /mnt/cache/Multimedia/Kinder/TVShows/Anna und die wilden Tiere/Anna und die wilden Tiere.S01E21.Wo pfeift das Murmeltier_.mp4
May 17 19:39:51 deheinServer move: move_object: /mnt/cache/Multimedia/Kinder/TVShows/Anna und die wilden Tiere/Anna und die wilden Tiere.S01E21.Wo pfeift das Murmeltier_.mp4 File exists
May 17 19:39:51 deheinServer root: mover: finished