Very slow transfer speed Trunas scale 22.12.2 with SMB

[sretalla]

Also, it disconnected before the test could complete... what kind of network is the client on?

 

[sretalla]

OK, so your internet is 100Mbits (almost), but your LAN should usually be much faster (Gigabit).

 

[saebaryo]

That's not even 100Mbits... that will need to be investigated if you're going to hope for faster transfers.

Is this speed between my server and local machine right?
So, something wrong in my network set up then?

 

[sretalla]

You haven't described your hardware properly. Please look at the forum rules (link above in red) and post a more complete description of your setup (not screenshots).

 

[sretalla]

Is this speed between my server and local machine right?
So, something wrong in my network set up then?

That's what it looks like to me.

 

[saebaryo]

Also, it disconnected before the test could complete... what kind of network is the client on?

Using a Wifi 5GHZ network that is about 57% of range. Will try again with Ethernet now

 

[saebaryo]

You haven't described your hardware properly. Please look at the forum rules (link above in red) and post a more complete description of your setup (not screenshots).

Hardware is:

• Motherboard maker and model: I do not have the exact model but it is from an HP EliteDesk 800 G1 TWR
• CPU make and model: i3-4160 CPU @ 3.60GHz
• RAM quantity: 16GB DDR 3
• Hard drives, quantity, model numbers, and RAID configuration, including boot drives: 5 drives: 1 HDD 320GB Seagate 7200RPM, for Pool 1: 2 X Seagate 2TB 5900RPM, Pool 2: 1 Seagate 2TB 5900RPM and 1 Hitachi 2TB 7200RPM
• Hard disk controllers: Where can I find that? in Bios maybe?
• Network cards: Gigabit Intel i217 LM (Rev4)

Thanks!

 

[saebaryo]

Tried iperf with direct ethernet cable from the router:

Code:

Client connecting to 192.168.1.29, TCP port 5201
TCP window size: 85.0 KByte (default)
------------------------------------------------------------
[  1] local 192.168.1.162 port 55428 connected with 192.168.1.29 port 5201
tcp write failed: Connection reset by peer
shutdown failed: Transport endpoint is not connected
[ ID] Interval       Transfer     Bandwidth
[  1] 0.0000-0.2897 sec  34.2 MBytes   992 Mbits/sec

 

[sretalla]

992 Mbits/sec

That's more like it. almost 1 Gbit. What took you over 4 seconds on WiFi took 0.2 seconds on ethernet cable.

However, you also got another connection failure during the test.

can you get the output from dmesg on the server to see if the driver is giving any feedback about that?

 

[sretalla]

Hard disk controllers: Where can I find that? in Bios maybe?

It's likely to be directly from the chipset (Intel Q87 Express, according to this: https://support.hp.com/us-en/document/c03808397)

I would suggest that's probably OK enough not to be the bottleneck here.

Seems some issue with the NIC from what we've seen so far.

 

[saebaryo]

That's more like it. almost 1 Gbit. What took you over 4 seconds on WiFi took 0.2 seconds on ethernet cable.

However, you also got another connection failure during the test.

can you get the output from dmesg on the server to see if the driver is giving any feedback about that?

Attached the result, thanks a mill!

 

[sretalla]

No specific reports from the NIC other than the standard Kubernetes stuff... maybe consider stopping all your apps while doing this testing.

It's using the e1000e driver and I see a reference to it being a 2.5 Gig card... some memory of those early 2.5 Gbit models being really buggy.

Not sure if anyone with direct experience with one of those can chime in on success with using it under SCALE?

Do you have a different client to test from? and/or another network card/free slot in the server?

 

[sretalla]

I see a reference to it being a 2.5 Gig card... some memory of those early 2.5 Gbit models being really buggy.

Scratch that, was mis-reading the line above it.

 

[saebaryo]

No specific reports from the NIC other than the standard Kubernetes stuff... maybe consider stopping all your apps while doing this testing.

It's using the e1000e driver and I see a reference to it being a 2.5 Gig card... some memory of those early 2.5 Gbit models being really buggy.

Not sure if anyone with direct experience with one of those can chime in on success with using it under SCALE?

Do you have a different client to test from? and/or another network card/free slot in the server?

This the test without any app running.

For different client's test you mean iperf test?
Sure, I have I can try.

As for other NIC I only have those USB to Ethernet adapters that I doubt would be good wdyt?

 

[sretalla]

For different client's test you mean iperf test?

Yes, looking to eliminate that the disconnection problem is coming from that client in particular. If we can get a test to complete with the server, the server is good.

As for other NIC I only have those USB to Ethernet adapters that I doubt would be good wdyt?

Possibly OK for the purposes of eliminating the onboard NIC as a source of the error.

Depends on how confident you are with swapping over to using another NIC (or managing 2 in the same box properly)

 

[saebaryo]

New iperf test from another machine connected to the same network via Ethernet:

Code:

iperf -c 192.
168.1.29 -p 5201               
--------------------------------
----------------------------
Client connecting to 192.168.1.2
9, TCP port 5201
TCP window size: 85.0 KByte (def
ault)
--------------------------------
----------------------------
[  3] local 192.168.1.81 port 54
708 connected with 192.168.1.29
port 5201
write failed: Connection reset b
y peer
shutdown failed: Transport endpo
int is not connected
[ ID] Interval       Transfer
  Bandwidth
[  3] 0.0000-0.2877 sec  33.9 MB
ytes   990 Mbits/sec

 

[saebaryo]

Yes, looking to eliminate that the disconnection problem is coming from that client in particular. If we can get a test to complete with the server, the server is good.


Possibly OK for the purposes of eliminating the onboard NIC as a source of the error.

Depends on how confident you are with swapping over to using another NIC (or managing 2 in the same box properly)

Well I can add a Gigabit USB ethernet adapter to the server for testing purposes if needed.

 

[sretalla]

Well I can add a Gigabit USB ethernet adapter to the server for testing purposes if needed.

Given the same failure from that different client, it's looking more like the server NIC.

Please try it with the USB adapter instead of the onboard one and see if you get the same or different.

 

[saebaryo]

Given the same failure from that different client, it's looking more like the server NIC.

Please try it with the USB adapter instead of the onboard one and see if you get the same or different.

I tried to set up the USB NIC from the console or from the Trunas GUI, I can't.
I name it then set ipv4 and 6 dhcp to No, then in the alias I added 192.168.1.39/24 but it tell that 192.168.1.0/24 is in use.
What am I missing?

 

[sretalla]

You can't be using both NICs at the same time with the same IP.

You need to remove the IP from the first one, save (but don't test yet), then assign it to the other, save and test.