Large cluster of 45 nodes

Kosh

Active Member
Dec 24, 2019
79
5
28
44
Hi all
I read topics about large clusters and did not find a solution for myself. We have one of the clusters with 46 servers, and after about 36 servers, problems with the cluster began with active use (vm migration)
version of all nodes 7.3.3
1- corosync network is separated into a separate vlan/corosync is on the same subnet
2- knet_transport is used: sctp (the cluster feels more stable on it)

Code:
quorum {
  provider: corosync_votequorum
}
    
totem {
  cluster_name: Cluster
  config_version: 70
  interface {
    knet_transport: sctp
    linknumber: 0
  }
  ip_version: ipv4-6
  knet_ping_interval: 200
  knet_ping_timeout: 5000
  knet_pong_count: 1
  link_mode: passive
  secauth: on
  version: 2
}

is there anything else that can be tweaked to make the cluster feel normal? I do not want to split into several clusters



on the screenshots, its standard behavior at any load
after the end of the unsuccessful migration, the cluster is going back
1.JPG2.JPG
 
Hi all
I read topics about large clusters and did not find a solution for myself. We have one of the clusters with 46 servers, and after about 36 servers, problems with the cluster began with active use (vm migration)
version of all nodes 7.3.3
1- corosync network is separated into a separate vlan/corosync is on the same subnet
2- knet_transport is used: sctp (the cluster feels more stable on it)

Code:
quorum {
  provider: corosync_votequorum
}
   
totem {
  cluster_name: Cluster
  config_version: 70
  interface {
    knet_transport: sctp
    linknumber: 0
  }
  ip_version: ipv4-6
  knet_ping_interval: 200
  knet_ping_timeout: 5000
  knet_pong_count: 1
  link_mode: passive
  secauth: on
  version: 2
}

is there anything else that can be tweaked to make the cluster feel normal? I do not want to split into several clusters



on the screenshots, its standard behavior at any load
after the end of the unsuccessful migration, the cluster is going back
View attachment 47400View attachment 47401
Hi,
so if I understand you correctly corosync and storage are using the same physical network, and as soon as there is load on that network you loose quorum in the cluster, correct?

corosync requires a low latency network, so as soon as the latency gets to high you will loose quorum. The recommendation is to use a separately physical network just for the cluster network.
 

About

The Proxmox community has been around for many years and offers help and support for Proxmox VE, Proxmox Backup Server, and Proxmox Mail Gateway.
We think our community is one of the best thanks to people like you!

Get your subscription!

The Proxmox team works very hard to make sure you are running the best software and getting stable updates and security enhancements, as well as quick enterprise support. Tens of thousands of happy customers have a Proxmox subscription. Get yours easily in our online shop.

Buy now!