[sheepdog-users] some nodes not join to sheepdog cluster when using corosync.

[Andrew J. Hobbs]

This is a recurring theme.  Would it be advisable to add some documentation to the website recommending zookeeper for production, and corosync only for small scale testing?

________________________________________
From: sheepdog-users-bounces at lists.wpkg.org [sheepdog-users-bounces at lists.wpkg.org] on behalf of Liu Yuan [namei.unix at gmail.com]
Sent: Friday, May 23, 2014 1:28 AM
To: Saeki Masaki
Cc: sheepdog-users at lists.wpkg.org
Subject: Re: [sheepdog-users] some nodes not join to sheepdog cluster when using corosync.

On Fri, May 23, 2014 at 02:06:19PM +0900, Saeki Masaki wrote:
> Hi,
>
> When trying to build a Sheepdog cluster of 12 nodes using Corosync, we found a strange behavior.
> Launch a Sheep process on each server, some nodes were not join to the cluster.( for more infomation below )
> Which "node" or not , was unspecified.
>
> I'm not familiar with corosync, but corosync logged "enabling flow control" .
> because send message buffer is full .

Corosync is said to support small cluster (less than 10 nodes). So in your case,
I suggest you try zookeeper, which support thousands of nodes.

Thanks
Yuan
--
sheepdog-users mailing lists
sheepdog-users at lists.wpkg.org
http://lists.wpkg.org/mailman/listinfo/sheepdog-users