Jumbo frame size set - netmtu in COROSYNC.CONF ?

[leex12]

Not sure if I was snow blind to them in v7 but since upgrade to v8 I have tons of messages 'complaining' about MTU size.

Historically I have had MTU set on NICs, bridges and vlans set to 9000.

Since seeing these messages I have validated 8745 seems to be a sweet spot working between nodes, VM, LXc and desktop clients using ping

ping -M do -s 8745 pve06

Changed my MTU references to 8745 but the error didn't go away .. so I revised down and still didn't go away

Jun 16 12:16:56 pve01 corosync[1198]: [KNET ] pmtud: Starting PMTUD for host: 6 link: 1
Jun 16 12:16:56 pve01 corosync[1198]: [KNET ] udp: detected kernel MTU: 8517
Jun 16 12:16:56 pve01 corosync[1198]: [KNET ] pmtud: PMTUD completed for host: 6 link: 1 current link mtu: 8405
Jun 16 12:16:56 pve01 corosync[1198]: [KNET ] pmtud: Starting PMTUD for host: 3 link: 0
Jun 16 12:16:56 pve01 corosync[1198]: [KNET ] udp: detected kernel MTU: 8517
Jun 16 12:16:56 pve01 corosync[1198]: [KNET ] pmtud: PMTUD completed for host: 3 link: 0 current link mtu: 8405
Jun 16 12:16:56 pve01 corosync[1198]: [KNET ] pmtud: Starting PMTUD for host: 3 link: 1
Jun 16 12:16:56 pve01 corosync[1198]: [KNET ] udp: detected kernel MTU: 8517
Jun 16 12:16:56 pve01 corosync[1198]: [KNET ] pmtud: PMTUD completed for host: 3 link: 1 current link mtu: 8405
Jun 16 12:16:56 pve01 corosync[1198]: [KNET ] pmtud: Starting PMTUD for host: 1 link: 0
Jun 16 12:16:56 pve01 corosync[1198]: [KNET ] udp: detected kernel MTU: 8517

I have left the NICs set to 8745 but now have the bridges etc set to 8517. Having reduced the MTU size a couple of times and it not resolving the issue I feel I changing the wrong thing.

Are people setting the netmtu to force the MTU size? I currently don't have this entry in my corosync.conf

 

[mgabriel]

AFAIK, the `netmtu` setting has no effect in Corosync 3 using knet as the transport (Which is default since Corosync 3 / Proxmox VE 6). Can't tell why you're getting messages since v8.

Would you mind posting your corosync.conf? Did you try removing the netmtu setting from the config?

 

[leex12]

my question was shouls i be setting netmtu to stop the messages!

There is nothing exciting in my config

logging {
  debug: on
  to_syslog: yes
}

nodelist {
  node {
    name: pve01
    nodeid: 5
    quorum_votes: 1
    ring0_addr: 192.168.20.1
    ring1_addr: 10.107.0.1
  }
  node {
    name: pve02
    nodeid: 4
    quorum_votes: 1
    ring0_addr: 192.168.20.2
    ring1_addr: 10.107.0.2
  }
  node {
    name: pve03
    nodeid: 2
    quorum_votes: 1
    ring0_addr: 192.168.20.3
    ring1_addr: 10.107.0.3
  }
  node {
    name: pve04
    nodeid: 1
    quorum_votes: 1
    ring0_addr: 192.168.20.4
    ring1_addr: 10.107.0.4
  }
  node {
    name: pve05
    nodeid: 3
    quorum_votes: 1
    ring0_addr: 192.168.20.5
    ring1_addr: 10.107.0.5
  }
  node {
    name: pve06
    nodeid: 6
    quorum_votes: 1
    ring0_addr: 192.168.20.6
    ring1_addr: 10.107.0.6
  }
}

quorum {
  provider: corosync_votequorum
}

totem {
  cluster_name: x12
  config_version: 27
  interface {
    linknumber: 0
  }
  interface {
    linknumber: 1
  }
  ip_version: ipv4-6
  link_mode: passive
  secauth: on
  version: 2
}

 

[leex12]

i have moved everything back to 8745

Jun 18 18:38:44 pve01 corosync[1262]:   [KNET  ] pmtud: PMTUD completed for host: 2 link: 0 current link mtu: 8629
Jun 18 18:38:44 pve01 corosync[1262]:   [KNET  ] pmtud: Starting PMTUD for host: 2 link: 1
Jun 18 18:38:44 pve01 corosync[1262]:   [KNET  ] udp: detected kernel MTU: 8745

just seems like what ever I set the mtu to it wants it to be a hundred or so smaller.

 

[mgabriel]

Are the messages still there if you set debug to off? That's the only thing I see related to configuration.