Hi!

 

After having my main UnRAID fileserver successfully up and running for the past few month, I finally finished setting up my Backup-UnRAID machine which is meant to be moved off-site and connected via OpenVPN.

 

While my main server runs buttery smooth with full Gigabit Read/Write, I am only getting a max. of 60 MB/s to and from my new Backup-UnRAID server. No parity (yet), no cache, direct read and write from and to the array.

 

I've switched around cables and switches and tested from 5+ clients, all of them with the same result:

Stable 100 MB/s on the main server, a max. of 60 MB/s on the new one - be it a single big file or fewer smaller ones.

 

I'm a tad out of ideas at this point... Might anyone have an idea where I could look out for the problem?

 

The diagnostics file is attached... Thanks in advance! :-)

backupserver-diagnostics-20210223-1501.zip

Start by doing a single stream iperf test.

Hi Jorge!

 

I just did that between the two UnRAID servers, reaching a bitrate of around 950Mbits/sec., which looks pretty good and normal.

Yes it does, do get the same 60MB/s from the various drives?

Naturally, this with also my next task. 🙂

 

I ran the DiskSpeed plugin with a short test on all disks.

While the two slowest disk have a minimum speed of 77 and 83 MB/s, all other disks are at 100 MB/s and well above...

Doesn't make much sense, if it's not the network you shouldn't get the same speed with all devices, if possible try with a different source PC.

I already tested the read and write procedures from 5 different clients, via LAN as well as via Wi-Fi, all with the same results of a pretty much stable max 60 MB/s read/write. 😞

That suggests a network issue, though strange if iperf is normal.

I went on to test different switches, different cables and even isolated the server as well as different clients completely from the rest of the network, sadly without any improvements. I'm still capped at around 60 MB/s and out of ideas... 😞

And the iperf test was single stream like mentioned? If yes see how the server performs with internal transfers, best to disable parity first if not that can be the bottleneck.