[ClusterLabs] corosync-cfgtool -s shows all links not connected for one particular node

[Jan Friesse]

Dirk,

On 23/05/2022 19:02, Dirk Gassen wrote:
> Greetings,
> 
> I have a four-node cluster on Ubuntu Focal with the following versions:
>> libknet1: 1.15-1ubuntu1
>> corosync: 3.0.3-2ubuntu2.1
>> pacemaker: 2.0.3-3ubuntu4.3

3.0.3 corosync-cfgtool was buggy - basically first version with 
correctly working `cfgtool -a` is 3.1.5. If possible I would recommend 
to update (either compile from source or use proxmox repo 
http://download.proxmox.com/debian/pve/.

Honza

> 
> Each node is connected to two networks:
>> testras1:
>>   eth0  10.1.8.24/26
>>   eth1  192.168.21.227/24
>> testras2:
>>   eth0  10.1.8.25/26   eth1  192.168.21.119/24
>> testras3:
>>   eth0  10.1.8.66/26
>>   eth1  192.168.21.13/24
>> testras4:
>>   eth0  10.1.8.77/26   eth1  192.168.21.19/24
> 
> The totem section of corosync.conf on all nodes:
>> totem {
>>     version: 2
>>     cluster_name: BERND-RAS
>>     # Disable encryption
>>     secauth: off
>>     interface {
>>         linknumber: 0
>>         #knet_transport: udp|sctp
>>         #knet_link_priority: 0
>>     }
>>     interface {
>>         linknumber: 1
>>         #knet_transport: udp|sctp
>>         #knet_link_priority: 1
>>     }
>>     transport: knet
>> }
> and the nodelist section:
>> nodelist {     node {
>>             ring0_addr: 192.168.21.227
>>             ring1_addr: 10.1.8.24
>>             nodeid: 2036952047
>>             name: testras1
>>     }
>>     node {
>>             ring0_addr: 192.168.21.119
>>             ring1_addr: 10.1.8.25
>>             nodeid: 2036951939
>>             name: testras2
>>     }
>>     node {
>>             ring0_addr: 192.168.21.13
>>             ring1_addr: 10.1.8.66
>>             nodeid: 1921682113
>>             name: testras3
>>     }
>>     node {
>>             ring0_addr: 192.168.21.19
>>             ring1_addr: 10.1.8.77
>>             nodeid: 1921682119
>>             name: testras4
>>     }
>> }
> 
> On all nodes crm_mon shows all four nodes online:
>> Node List:
>>   * Online: [ testras1 testras2 testras3 testras4 ]
> and "corosync-cfgtool -s" shows the very same:
>> Printing link status.
>> Local node ID 2036952047
>> LINK ID 0
>>     addr    = 192.168.21.227
>>     status:
>>         nodeid 1921682113:    link enabled:1    link connected:1
>>         nodeid 1921682119:    link enabled:1    link connected:1
>>         nodeid 2036951939:    link enabled:1    link connected:1
>>         nodeid 2036952047:    link enabled:1    link connected:1
>> LINK ID 1
>>     addr    = 10.1.8.24
>>     status:
>>         nodeid 1921682113:    link enabled:1    link connected:1
>>         nodeid 1921682119:    link enabled:0    link connected:1
>>         nodeid 2036951939:    link enabled:1    link connected:1
>>         nodeid 2036952047:    link enabled:1    link connected:1
> 
> 
> However, when I add a node that doesn't exist that changes:
>>     node {
>>             ring0_addr: 192.168.120.13
>>             ring1_addr: 10.1.8.99
>>             nodeid: 2036942833
>>             name: testras5
>>     }
> Now "corosync-cfgtool -s" shows:
>> Printing link status.
>> Local node ID 2036952047
>> LINK ID 0
>>     addr    = 192.168.21.227
>>     status:
>>         nodeid 1921682113:    link enabled:1    link connected:0
>>         nodeid 1921682119:    link enabled:1    link connected:1
>>         nodeid 2036942833:    link enabled:1    link connected:1
>>         nodeid 2036951939:    link enabled:1    link connected:1
>>         nodeid 2036952047:    link enabled:1    link connected:1
>> LINK ID 1
>>     addr    = 10.1.8.24
>>     status:
>>         nodeid 1921682113:    link enabled:1    link connected:0
>>         nodeid 1921682119:    link enabled:1    link connected:1
>>         nodeid 2036942833:    link enabled:0    link connected:1
>>         nodeid 2036951939:    link enabled:1    link connected:1
>>         nodeid 2036952047:    link enabled:1    link connected:1
> while everything else stays the same.
> 
> Why would "link connected" show 0 for one of the existing nodes but not 
> for the non-existing node (2036942833)? (All existing nodes can still 
> see each other) What am I missing?
> 
> Dirk
> _______________________________________________
> Manage your subscription:
> https://lists.clusterlabs.org/mailman/listinfo/users
> 
> ClusterLabs home: https://www.clusterlabs.org/
>