MTU Warning

[np-prxmx]

Hi ALL,
i've some nodes on OVH
Every node has 2 physical network interfaces

1. is not support MTU 9000 --> vmbr1
2. is support MTU 9000 (vRACK) --> vmbr0

I've setup my cluster, where link 0 is on vmbr0 (for every node) and link 1 on vmbr1 (for every node). I've also insert in corosync.conf the setting "netmtu : 1500 " to force communication between nodes only with MTU 1500, because one interface doesn't support it.

But, i obtain this error :

Nov 24 12:15:11 NODE4 corosync[4256]: [KNET ] pmtud: possible MTU misconfiguration detected. kernel is reporting MTU: 8988 bytes for host 2 link 1 but the other node is not acknowledging packets of this size.
Nov 24 12:15:11 NODE4 corosync[4256]: [KNET ] pmtud: This can be caused by this node interface MTU too big or a network device that does not support or has been misconfigured to manage MTU of this size, or packet loss. knet will continue to run but performances might be affected.

I suppose that "link 1" refers to network vmbr1 so, the interface where i haven't support to MTU 9000, but i can't force it to use MTU 1500.

Any ideas?
Thanks in advance!

 

[spirit]

'I have also insert in corosync.conf the setting "netmtu : 1500 "

do you also have increase config_version ? because the new config will not be applied without this.

I suppose that "link 1" refers to network vmbr1

link0 or 1, use ip addresses defined in /etc/pve/corosync.conf on nodes


for example

  node {
    name: nodename1
    nodeid: 1
    quorum_votes: 1
    ring0_addr: 192.168.0.1   #this is link0
    ring1_addr: 10.0.0.1 #thiis is link1
  }

 

[np-prxmx]

do you also have increase config_version ? because the new config will not be applied without this.

Yes, of course. My version number is 5, and all nodes have it, and it was replicated and in /etc/corosync/corosync.conf correctly byself.

 

[spirit]

I don't known if you can create vlan interface inside the vrack, but if yes, you could have a dedicated vlan interface with a lower mtu than 9000.

something like

auto vmbr0
iface vmbr0 inet manual
     bridge-ports eth0
     mtu 9000

auto eth0.1000
iface eth0.1000 inet static
     address ....
     mtu 1500

 

[np-prxmx]

Yes, i can create vlan.

But, don't understand your suggestion.

I've this conf


auto eth0
iface eth0 inet manual
mtu 9000
#vRACK - In USE

auto eth1
iface eth1 inet manual
#Public - In USE


auto vmbr0
iface vmbr0 inet static
address 1.1.1.1/24
gateway 10.10.10.10
bridge-ports eth0
bridge-stp off
bridge-fd 0
mtu 9000
#MGMT vRack

auto vmbr1
iface vmbr1 inet static
address 11.11.11.11
bridge-ports eth1
bridge-stp off
bridge-fd 0
#MGMT Public


root@:~# cat /etc/pve/corosync.conf
logging {
debug: off
to_syslog: yes
}

nodelist {
node {
name: NODE1
nodeid: 1
quorum_votes: 1
ring0_addr: 1.1.1.1
ring1_addr: 11.11.11.11
}
node {
name: NODE2
nodeid: 2
quorum_votes: 1
ring0_addr: 1.1.1.2
ring1_addr: 11.11.11.12
}
node {
name: NODE3
nodeid: 3
quorum_votes: 1
ring0_addr: 1.1.1.3
ring1_addr: 11.11.11.13
}
node {
name: NODE4
nodeid: 4
quorum_votes: 1
ring0_addr: 1.1.1.4
ring1_addr: 11.11.11.14
}
}

quorum {
provider: corosync_votequorum
}

totem {
cluster_name: CL01
config_version: 5
interface {
linknumber: 0
}
interface {
linknumber: 1
}
ip_version: ipv4-6
link_mode: passive
secauth: on
version: 2
netmtu: 1500
}

 

[spirit]

I mean instead of using netmtu in corosync.conf,

setup mtu directly on a vlan interface for ip on vrack.

for example, something like:

auto eth0
iface eth0 inet manual
mtu 9000
#vRACK - In USE

auto eth0.1000
auto eth0.1000 inet static
    address 1.1.1.1/24
    gateway 10.10.10.10
    mtu 1500
   #MGMT vRack

auto eth1
iface eth1 inet manual
#Public - In USE


auto vmbr0
iface vmbr0 inet manual
   bridge-ports eth0
   bridge-stp off
   bridge-fd 0
   mtu 9000


auto vmbr1
iface vmbr1 inet static
address 11.11.11.11
bridge-ports eth1
bridge-stp off
bridge-fd 0
#MGMT Public

 

[np-prxmx]

I've already done it.

 

[np-prxmx]

The error is ever reported on interface that hasn't got MTU set to 9000, but i'd insert the interface in cluster configuration as failover. For this i've setup mtu in corosync, because so i theorically, i forced calls between nodes in 1500.

 

[np-prxmx]

Hello,
somebody with same problem?