repeatedly problem with proxmox.

[Melanxolik]

We have small cluster with 15 nodes.
This cluster always has problem cluster syncronization, we always need restart cluster nodes.
Proxmox very does not stable product on cluster systems.

/etc/init.d/pvestatd does not fix problem syncronization.

# clustat 
Cluster Status for freu @ Thu Jun 11 11:38:21 2015
Member Status: Quorate

 Member Name                                                     ID   Status
 ------ ----                                                     ---- ------
 cluster-1-1                                                         1 Online, Local, rgmanager
 cluster-1-2                                                         2 Online, rgmanager
 cluster-1-3                                                         3 Online, rgmanager
 cluster-1-4                                                         4 Online, rgmanager
 cluster-1-5                                                         5 Online, rgmanager
 cluster-1-6                                                         6 Online, rgmanager
 cluster-1-7                                                         7 Online, rgmanager
 cluster-1-8                                                         8 Online, rgmanager
 cluster-1-9                                                         9 Online, rgmanager
 cluster-1-10                                                       10 Online, rgmanager
 cluster-1-11                                                       11 Online, rgmanager
 cluster-1-12                                                       12 Online, rgmanager
 cluster-1-13                                                       13 Online, rgmanager
 cluster-1-14                                                       14 Online, rgmanager
 cluster-1-15                                                       15 Online, rgmanager

 Service Name                                                     Owner (Last)                                                     State         
 ------- ----                                                     ----- ------                                                     -----         
 service:web-adm                                                  cluster-1-14                                                     started

# group_tool ls
fence domain
member count  15
victim count  0
victim now    0
master nodeid 1
wait state    none
members       1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 

dlm lockspaces
name          rgmanager
id            0x5231f3eb
flags         0x00000000 
change        member 15 joined 1 remove 0 failed 0 seq 23,23
members       1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 

name          clvmd
id            0x4104eefa
flags         0x00000000 
change        member 15 joined 1 remove 0 failed 0 seq 26,26
members       1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

this problem appeared after two reboot one node in cluster and such a problem when we use constantly proxmox

 

[nethfel]

Now I could be very wrong about this - but AFAIK, you can't have community and non-community in the same cluster, and all systems in a cluster have to be under the same level of support... I wonder if that is causing some of your problems? Possibly different versions of software available via subscription vs community edition

 

[Melanxolik]

You don't right, all nodes in cluster have:

Proxmox VE Community Subscription 1 CPU/year

but, cluster does't very stable work last year.


​

 

[nethfel]

Ok, I wasn't sure if all nodes had the subscription or not as the picture you posted only showed the two green nodes which made me wonder if the others were not on subscription.

 

[fireon]

Downgrade to kernel 2.6.32-34-pve. And test it should be work. We had the same problem, with every kernel newer then 2.6.32-34-pve. We don't know why, maybe the High Number of VMs. And we have the same problem on to different locations. Network is on both ok. I also tested it with omping, and when the cluster ist broken omping works fine! So, i don't know.... Bug? ...

 

[Melanxolik]

how does storage connected to cluster? me interested topology connect.