Problem when activate Jumbo Frames on my network

[ViMx]

Hi,

I encounter a problem when I'm trying to activate jumbo frames on my network.

My installation:
- Synology DS411+ (MTU @9000)
- Ubiquiti ERL (MTU WAN inteface set to 9000)
- Compatible switchs (Netgear GS608V2 + GS105)
- My desktop W10 (MTU set @9000)

The problem is when I activate the jumbo frames on all my devices, I'm unable to access the DSM via my domain name. No problem to access it via the local IP adress, via SMB or FTP (performance is good !)

Here is the result when I try to access any Web service on my NAS:

With the domain:


With the local IP:


Obviously the packets drops with when I use the domain... As soon as I deactivate the jumbo frames (Either on the Syno or on the Pc), no problem. I don't have any problem either when I'm trying to access it with my smartphone (Either with the domain or the IP, via Wifi or 4G)

Any idea to fix this ?

Thanks.

 

[netwrks]

Remove Jumbo frame off of WAN. You won't see a benefit.. What about the Ethernet port(s) on the ERL to your switch(es)? Did you enable jumbo frame there?

 

[ViMx]

Hi netwrks

Thanks for your help.

My ERL config:

eth0 is connected to switchs
eth1 is connected to the ONT of my provider
eth2 is connected to my triple play box (For TOIP and TVIP)

(I've missed used "WAN interface". I obviously meant "LAN interface")

 

[netwrks]

So your switch port for your NAS and all uplinks on your switches all have jumbo frames enabled? You may have to reboot your gear once you configured it. What's br0 for on the ERL?

 

[ViMx]

So your switch port for your NAS and all uplinks on your switches all have jumbo frames enabled? You may have to reboot your gear once you configured it. What's br0 for on the ERL?

Yes, all have JF are activated since:

C:\Users\User>ping 192.168.1.2 -f -l 9000

Envoi d’une requête 'Ping' 192.168.1.2 avec 9000 octets de données :
Réponse de 192.168.1.2 : octets=9000 temps=1 ms TTL=64
Réponse de 192.168.1.2 : octets=9000 temps=1 ms TTL=64
Réponse de 192.168.1.2 : octets=9000 temps=1 ms TTL=64
Réponse de 192.168.1.2 : octets=9000 temps=1 ms TTL=64

Statistiques Ping pour 192.168.1.2:
Paquets : envoyés = 4, reçus = 4, perdus = 0 (perte 0%),
Durée approximative des boucles en millisecondes :
Minimum = 1ms, Maximum = 1ms, Moyenne = 1ms

And I've tested a reboot

 

[ViMx]

This is really weird...

I've tested by editing my hosts and it works (But I have a translation port problem now -_- )... so it seems that's the loopback + jumbo frame combo that causes this...

 

[thiggins]

This is really weird...

I've tested by editing my hosts and it works (But I have a translation port problem now -_- )... so it seems that's the loopback + jumbo frame combo that causes this...

I'd be interested to know if Jumbo frames improve performance. They are a holdover from an earlier time when bus bandwidths and Ethernet PHYs didn't have offload processors...

 

[sfx2000]

Makes no sense to me - jumbo's are typically negotiated...

Something else is wrong, and it's likely not jumbo frames specifically... maybe a driver issue perhaps...