+- PINE64 (https://forum.pine64.org)
+-- Forum: ROCKPRO64 (https://forum.pine64.org/forumdisplay.php?fid=98)
+--- Forum: Linux on RockPro64 (https://forum.pine64.org/forumdisplay.php?fid=101)
+--- Thread: Please help me troubleshoot strange network performance (/showthread.php?tid=11167)
Please help me troubleshoot strange network performance - UnoqueVa - 08-27-2020
Hi, first post here!
My big project this summer was to build a NAS with a RockPro64 at its heart, making use of its PCIe connector to connect an LSI SAS HBA and a SAS expander to connect much more than 4 drives. On top of that, I was able to install ZFS quite easily and use it as a remote backup server for my main NAS. I don't thinks there's many other SBCs than can handle all of that.
After a lot of reading around here and Github, I only tried installing ayufan's pre-release from April, because it added the neccessary support for the LSI card. I chose the OMV version buster-openmediavault-rockpro64-0.10.12-1184-arm64.img.xz and things have been more or less smooth (my gripes are about OMV more than the RP64).
Now, the problem is that I'm trying to send backups between my servers and I'm hitting a strange 300 KByte/s bottleneck when sending from the remote RP64 to my local proxmox Ryzen server. The same test in the other direction achieves about 30 MByte/s (both locations have 300 Mbit fiber internet connections).
Next I've tried with iperf and a Mac Mini that is in the same location as the RockPro64 and the results stump me again:
Code:
Proxmox 22 MBytes/s ----(cloud)---> Mac Mini
10~20 MBytes/s <--(_______)---
Proxmox 35 MBytes/s -----(cloud)---> RockPro64
0.37 MBytes/s <---(_______)---
RockPro64 ~30 MBytes/s -----LAN---> Mac Mini
20~40 MBytes/s <----------So it seems like the RockPro is capable of sending 30 MByte/s upsteam its network port, just not towards my Proxmox server.
And the router can send me a reasonably fast stream, just not from my RockPro64
I've tested raw iperf3, zfs sends through ssh, zfs sends through Wireguard... the 300 Kb/s limit is consistent.
Please help me, what is it that I'm missing?
RE: Please help me troubleshoot strange network performance - guannais - 11-01-2020
With mainline kernel I had some problems with TCP connections. By inspecting them with Wireshark I saw that some IP packets were dropped and asked for retransmition likely due to bad checksum. The problems disappeared when I turned off "TCP checksum offload" from the network device:
Code:
# ethtool -K eth0 tx off