Slow transfer speed between two Truenas systems over SMB

DonZalmrol

Dabbler
Joined
Oct 6, 2021
Messages
14
Hi,

I have two Truenas systems (both 12.0-U7) with the following specs:

Main fileserver (runs 24x7):
  • HP DL380 G9
  • 1x Intel Xeon E5-2650L v3 @ 1.80GHz
  • 128GB DDR4 ECC
  • HP P440AR in HBA mode
  • 12G SAS expander
  • Storagepool with 5x WD 14TB Ultrastar DC HC530's in RaidZ2 (SATA3)
  • 10GB LACP, no jumbo, HP 560FLR-SFP+
Backup fileserver (which only runs a couple of hours a day):
  • Custom build Supromicro X10SRL-F
  • 1x Intel Xeon E5-2650 v3 @ 1.80GHz
  • 64GB DDR4 ECC
  • 2x LSI HBA's SAS2008 (6GB SAS2)
  • Storagepool with a mix of 12 Enterprise HDDs 4TB with SATA3 in RaidZ1
  • 10GB LACP, no jumbo, Intel 82599ES-SFP+
Core switch is a MikroTik CRS317-1G-16S+

When I do a replication job from main to backup the max speed is about 350MiB/s and throttles to min 150MiB/s. Instead of going to min or above 1GiB/s.

Speed from 10G switch (+- 2Gbps):
1643904463550.png


Main speed, CPU and MEM:
1643897035570.png
1643897985264.png


Backup speed, CPU and MEM:
1643897063899.png
1643898017289.png


What I notice is that the CPU of my main is 100% in use on thread #2, I believe this is either SMB or the network card.
Could It be that I need to configure my card to utilize more theads?

1643897174649.png


The spread over my backup is more equal:
1643897270350.png


What I tried:
  1. Followed this thread to enable the RSS -> https://www.truenas.com/community/t...trouble-shooting-approach-all-ssd-pool.98452/
  2. Enabled/ Disabled jumbo frames (MTU 9000)
  3. Added/ removed cache and log drives (no difference, running without now)
  4. Checked the switch
Any tips on improving the speeds? I'm quite sure the bottleneck is coming from my main server due to the one high thread (#2).

Thanks!
 
Last edited:
Top