the Ethernet interface is not the issue
using Samba everything works fine
the Ethernet interface is not the issue
using Samba everything works fine
the Ethernet interface is not the issue
using Samba everything works fine
Hi JurKub - agree - I don't think it is the ethernet card / cabling / ... itself either. Just trying to isolate the Kernel Settings / Drivers / ....
Given the number of changes all happening - think that "a regression" may be a possibility. I just don't know where to look as such. E.G is it sending / receiving / window size / UDP / TCP / ....
what does
``time cp /nfs/abc /ramdisk/
take ... etc
Guys thx for your passion
I believe it's an rsize / wsize issue of the Raspi NFS Server
Nevertheless what I enter on client site using mount for rsize and wsize the value don't change
For me it looks like it is hard coded into the Raspi NFS Server or I'm to stupid to find the way to change it
NFS mount (rsize=131072,wsize=131072)
10.0.0.53:/srv/vdr/video on /storage/videos type nfs (rw,relatime,vers=3,rsize=131072,wsize=131072,namlen=255,hard,proto=tcp,timeo=600,retrans=2,sec=sys,mountaddr=10.0.0.53,mountvers=3,mountproto=tcp,local_lock=none,addr=10.0.0.53)
Samba mount (rsize=1048576,wsize=1048576)
Based on https://github.com/torvalds/linux/search?q=RPCSVC_MAXPAYLOAD the kernel is set to 1MB for TCP
So what are the tcp window sizes? And the latency?
nuc11:~ # cat /proc/sys/net/ipv4/tcp_wmem
4096 16384 4194304
nuc11:~ # cat /proc/sys/net/ipv4/tcp_rmem
4096 131072 6291456
nuc11:~ # cat /proc/sys/net/ipv4/tcp_window_scaling
1
I believe it's an rsize / wsize issue of the Raspi NFS Server
Nevertheless what I enter on client site using mount for rsize and wsize the value don't change
did you read in the link to nfs from comment #15 the chapter "Transport Methodes" ?
to it seems UDP, TCP, wsize, rsize, MTU are depending on each other...
maybe a stupid hint: changing the the config file on the server needs a restart of the server
some more links:
some questions:
1. did your lan setup ever run or is it a regression ( I esp. mean: simultaneous recording and replaying) ?
2. you said running SMB is fine, why don't run SMB [3.1 ?!] only ?
3. why are you shoveling big data through an LAN ?
I esp. mean when all 4 satellites on the intel box are recording
why not store the data where they first appear (at the intel box) ?
4. what does the network traffic and the CPU load on an PI look like when doing 4x recording *and* 4x replaying concurrently (!) ?
did you checked it with htop, etc. ?
- I've no idea from PI or FHEM or VDR) -
5. but isn't the intel box a lot more potent then the PI and wouldn't it make more sense to run a NFS server on the intel box too and in the dependence of what OS is currently running on the intel to maybe use a newer version of NFS (4.x) / kernel ? [1]
[1]
newer mostly means "better", but the opposite might also be true (sometimes it is, but sometimes only !)
I guess you're nailed with an aged kernel on the PI ?
I've read newer kernels benefiting from more throughput ...
I've currently also no idea if NFS 4.x provides any benefit over 3.x, but usually a newer version fixes the bugs/shortcommings in an elder version ... (sometimes )
or I'm to stupid to find the way to change it
/etc/nfsmount.conf
see:
https://linux.die.net/man/5/nfsmount.conf
clear (?): adjusting {r,w}size => no need to provide them as mount option (client) - I would like to say - dito: Proto, etc.
===
btw.: google is also a search engine
Display MoreBased on https://github.com/torvalds/linux/search?q=RPCSVC_MAXPAYLOAD the kernel is set to 1MB for TCP
So what are the tcp window sizes? And the latency?
nuc11:~ # cat /proc/sys/net/ipv4/tcp_wmem
4096 16384 4194304
nuc11:~ # cat /proc/sys/net/ipv4/tcp_rmem
4096 131072 6291456
nuc11:~ # cat /proc/sys/net/ipv4/tcp_window_scaling
1
fhemraspi:~ # cat /proc/sys/net/ipv4/tcp_wmem
4096 16384 4194304
fhemraspi:~ # cat /proc/sys/net/ipv4/tcp_rmem
4096 131072 6291456
fhemraspi:~ # cat /proc/sys/net/ipv4/tcp_window_scaling
1
2. you said running SMB is fine, why don't run SMB [3.1 ?!] only ?
in the meantime I changed to Samba on all machines.
But I'll still try to find the NFS issue
But I'll still try to find the NFS issue
... and please report here