[prev in list] [next in list] [prev in thread] [next in thread] 

List:       gluster-devel
Subject:    [Gluster-devel] Glusterfs HA & AFR
From:       chris_mavin () hotmail ! com (Chris Mavin)
Date:       2009-02-24 11:09:35
Message-ID: BAY138-DS2B4597EDD12E0ABBCB9F38EAF0 () phx ! gbl
[Download RAW message or body]

Thank you for the clarification Daniel.

That is what I suspected.

-----Original Message-----
From: Daniel Maher [mailto:dma+gluster at witbe.net] 
Sent: 18 February 2009 09:41
To: Chris Mavin
Cc: gluster-devel at nongnu.org
Subject: Re: [Gluster-devel] Glusterfs HA & AFR

Chris Mavin wrote:

> Can someone clarify the following for me.
>  
> I'm using Glusterfs RC 2.
>  
> If I am using the HA translator rather than AFR, Does this expect to be 
> using a shared storage accessible from both servers?
>  
> Put another way, is the HA translator simply a way of providing High 
> availability for a SAN?
>  
> When I test it using 2 servers the active server is written to fine but 
> the backup server simply gets a 0 byte file written to it.

HA is, afaik, just a way to switch communications between target 
machines - it does not itself have any sort of replication features, nor 
does it have the ability to speak to multiple targets simultaneously.

Given three machines :
- ClientA
- ServerA
- ServerB

If ClientA is configured to communicate to ServerA & ServerB via HA, it 
will pick one (ServerA, for example), and speak to it exclusively until 
such a time as ServerA stops working - whereupon it will communicate 
with ServerB.

Nowhere in that scenario is data replicated or copied between ServerA 
and ServerB.  For this, you'd either need to set up server-side AFR 
between ServerA and ServerB (and keep HA on ClientA), or set up 
client-sde AFR on ClientA (and ditch HA altogether).

-- 
Daniel Maher <dma+gluster AT witbe DOT net>





[prev in list] [next in list] [prev in thread] [next in thread]