Server Installation 2021

[TerryJC]

I'll send it off list.

[TerryJC]

Okay, I'm in ...

That's a relief. I couldn't even get that far this morning.

[hamishmb]

The new results are below.

Bear in mind that these weren't taken connecting to WMT, instead I connected to Terry's network - the ping results, especially, may not be apples-to-apples.

Pinging minster Pi:

Code: Select all

PING 192.168.0.50 (192.168.0.50) 56(84) bytes of data.
64 bytes from 192.168.0.50: icmp_seq=1 ttl=63 time=23.0 ms
64 bytes from 192.168.0.50: icmp_seq=2 ttl=63 time=22.4 ms
64 bytes from 192.168.0.50: icmp_seq=3 ttl=63 time=23.0 ms
64 bytes from 192.168.0.50: icmp_seq=4 ttl=63 time=24.2 ms
64 bytes from 192.168.0.50: icmp_seq=5 ttl=63 time=25.2 ms

--- 192.168.0.50 ping statistics ---
5 packets transmitted, 5 received, 0% packet loss, time 4005ms
rtt min/avg/max/mdev = 22.352/23.531/25.163/1.010 ms

Better, but still not great.

speedtest-cli from Sump Pi:

Code: Select all

Retrieving speedtest.net server list...
Selecting best server based on ping...
Hosted by Lightning Fibre Ltd (London) [85.56 km]: 23.049 ms
Testing download speed................................................................................
Download: 55.07 Mbit/s
Testing upload speed......................................................................................................
Upload: 15.89 Mbit/s

Far better, but not from WMT's slow internet connection.

Transferring a 50MiB zero-filled file from and to the minster Pi (/tmp to avoid SD card limitations):

From minster Pi:

Code: Select all

zero.img                                      100%   50MB 985.6KB/s   00:51    

To minster Pi:

Code: Select all

zero.img                                      100%   50MB   1.1MB/s   00:45    

Not as bad, but still not clear what the slowdown is about. CPU load is much lower on the VPN server, so I'm not sure what the delay is about. Maybe single-core performance encrypting SSH traffic? What model of pi is the minster pi?

Note: The SSH connection to the VPN server stayed usable this time, so I could observe CPU load properly. General input lag was much less noticeable this time.

[hamishmb]

Another thing to test when we can is to see if multiple simultaneous users will work.

[TerryJC]

We can't really do that until we have the system installed at WMT. Whatever the result of that is it will be what we have. We can probably live with it if there is still a problem.

[TerryJC]

Thanks Hamish for your help here.

[TerryJC]

What model of pi is the minster pi?

A Pi 3 and the Minster Bells is a Pi Zero.

[TerryJC]

Hamish,

A thought. Did you do the MTU Thing?

[hamishmb]

Are you sure it's a Pi 0? It looked like it was running the armv7 kernel, which I pi zero won't be able to do.

I did do the MTU thing, but that shouldn't make _too_ huge a difference - eg it doesn't if I'm downloading large files after setting it.

[TerryJC]

Are you sure it's a Pi 0? It looked like it was running the armv7 kernel, which I pi zero won't be able to do.

I'm assuming that you only logged in to the Music Pi. That is a Pi 3. It's the Bells Pi that's a Zero.

I did do the MTU thing, but that shouldn't make _too_ huge a difference - eg it doesn't if I'm downloading large files after setting it.

I just recalled that we started doing that because we had a problem in the early days.

Thinking about it, didn't you do something on the Pis to set the MTU as well as at the Client end? None of these Pis will have that,