Functionally Split Cluster

Looking at the design of a functionally split cluster, I
have a few  
questions.
http://netwinsite.com/surgemail/help/clustering_fu
nctional.htm

For the front end "filter server," does each
server need to be  
configured manually when a domain is added? Or does the back
end  
server make any of those updates when a domain is added?
Assuming the  
front ends must be configured separately, is it possible to
set up so  
surgemail.ini can be sync'd between them? I believe there
are host  
specific settings

For domains in "surgewall" mode (incoming SMTP and
filtering only,  
with mail directed to an outside server), do accounts need
to be  
created on the main server so people can log in to manage
filtering?  
If it is kept local on the filtering server, how is spam
info shared  
between front end servers?

For the webmail servers, should I be planning to have each
domain go  
to a specific Webmail server so the cached data is
available? If so,  
is there a recommended way to sync between webmail servers
in case  
the person's primary is not available and they log in to a
different  
server? When the server is all on a single machine, adding a
domain  
also configures webmail. Is this possible with the servers
split, or  
do I need to manually update webmail.ini on the separate
machines?  
Can webmail.ini be sync'd between multiple webmail machines
(maybe  
even from an unused webmail configuration on the main
server), or are  
there host specific settings I need to be careful with?

I know I need to sync srs.secret among all the front/back
end servers  
(not webmail, though?). Are there other files that should be
sync'd  
periodically, or upon updates?

My apologies if any of this is explained in the
documentation or  
FAQs. I've reviewed the manual, but wanted to clarify the
items above  
before I try to set up a test system to verify exactly how
the  
functional split is managed.

Thanks!

Noah

Noah Price wrote:
> Looking at the design of a functionally split cluster,
I have a few 
> questions.
> http://netwinsite.com/surgemail/help/clustering_fu
nctional.htm
> 
> For the front end "filter server," does each
server need to be 
> configured manually when a domain is added? 

Yes

Or does the back end server
> make any of those updates when a domain is added?

no

  Assuming the front
> ends must be configured separately, is it possible to
set up so 
> surgemail.ini can be sync'd between them? 
I believe there are host
> specific settings
> 
> For domains in "surgewall" mode (incoming
SMTP and filtering only, with 
> mail directed to an outside server), do accounts need
to be created on 
> the main server so people can log in to manage
filtering? If it is kept 
> local on the filtering server, how is spam info shared
between front end 
> servers?

there must be one front end server per domain

> 
> For the webmail servers, should I be planning to have
each domain go to 
> a specific Webmail server so the cached data is
available?

It is probably best yes.


  If so, is
> there a recommended way to sync between webmail servers
in case the 
> person's primary is not available and they log in to a
different server? 
> When the server is all on a single machine,

That is automatic.

  adding a domain also
> configures webmail. Is this possible with the servers
split, or do I 
> need to manually update webmail.ini on the separate
machines? 

Yes you need to manually update webmail.ini

Can
> webmail.ini be sync'd between multiple webmail machines
(maybe even from 
> an unused webmail configuration on the main server), or
are there host 
> specific settings I need to be careful with?

I think you could probably sync the file itself.

> 
> I know I need to sync srs.secret among all the
front/back end servers 
> (not webmail, though?). Are there other files that
should be sync'd 
> periodically, or upon updates?

no

> 
> My apologies if any of this is explained in the
documentation or FAQs. 
> I've reviewed the manual, but wanted to clarify the
items above before I 
> try to set up a test system to verify exactly how the
functional split 
> is managed.
> 
> Thanks!
> 
> Noah
> 

A lot of these questions stem from building a cluster to
handle multiple 
domains and be functionaly split, we belive this is the
wrong approach, 
we believe you should split on domains, and only use a
functionaly split 
cluster to handle a single large domain which cannot be
contained on a 
single or mirrored server.

I'm not saying it won't work, it's just not the way I'd do
it 



-- 
When replying please quote original message to save time and
always let
us know what OS and version of SurgeMail you are using.

************************************************************
****************
If you wish to communicate with other users and talk about
SurgeMail and
also get news about updates etc then join the SurgeMail
List.
Send an email to surgemail-list-requestnetwinsite.com with only 
"subscribe" in the message body

FAQ - http
://www.netwinsite.com/surgemail/help/faq.htm
************************************************************
**************** 

Thanks -- a few more questions below.

On Dec 10, 2007, at 4:48 PM, SurgeMail Support wrote:

> Noah Price wrote:
>> Looking at the design of a functionally split
cluster, I have a  
>> few questions.
>> http://netwinsite.com/surgemail/help/clustering_fu
nctional.htm
>> For the front end "filter server," does
each server need to be  
>> configured manually when a domain is added?
>
> Yes
>
>> Or does the back end server make any of those
updates when a  
>> domain is added?
>
>
> no
>
>> Assuming the front
>> ends must be configured separately, is it possible
to set up so  
>> surgemail.ini can be sync'd between them?
> I believe there are host specific settings
>> For domains in "surgewall" mode (incoming
SMTP and filtering only,  
>> with mail directed to an outside server), do
accounts need to be  
>> created on the main server so people can log in to
manage  
>> filtering? If it is kept local on the filtering
server, how is  
>> spam info shared between front end servers?
>
> there must be one front end server per domain

Okay, it sounds like I should have a single server that is
primary  
for each domain, which is fine, and the other SMTP servers
would all  
relay to it (i.e., secondary MX). But for reliability, I'd
like to  
set up each function so there's a second live server that
can handle  
the function. Is there a recommended way to achieve that?

>> For the webmail servers, should I be planning to
have each domain  
>> go to a specific Webmail server so the cached data
is available?
>
> It is probably best yes.
>
>
>> If so, is
>> there a recommended way to sync between webmail
servers in case  
>> the person's primary is not available and they log
in to a  
>> different server? When the server is all on a
single machine,
>
> That is automatic.

Great -- is that sync'd automatically between webmail
servers, or do  
you mean the "new" webmail server will
automatically retrieve  
everything from the mail server?

>>  adding a domain also
>> configures webmail. Is this possible with the
servers split, or do  
>> I need to manually update webmail.ini on the
separate machines?
>
> Yes you need to manually update webmail.ini
>
>> Can
>> webmail.ini be sync'd between multiple webmail
machines (maybe  
>> even from an unused webmail configuration on the
main server), or  
>> are there host specific settings I need to be
careful with?
>
> I think you could probably sync the file itself.
>
>> I know I need to sync srs.secret among all the
front/back end  
>> servers (not webmail, though?). Are there other
files that should  
>> be sync'd periodically, or upon updates?
>
> no
>
>> My apologies if any of this is explained in the
documentation or  
>> FAQs. I've reviewed the manual, but wanted to
clarify the items  
>> above before I try to set up a test system to
verify exactly how  
>> the functional split is managed.
>> Thanks!
>> Noah
>
> A lot of these questions stem from building a cluster
to handle  
> multiple domains and be functionaly split, we belive
this is the  
> wrong approach, we believe you should split on domains,
and only  
> use a functionaly split cluster to handle a single
large domain  
> which cannot be contained on a single or mirrored
server.

Sure, I'm happy to split on domains, especially if that's
what you  
recommend   But I'd
like to have at least two servers to handle  
each domain for any given function in case of any sort of
equipment  
failure, which led to my questions about webmail. For
incoming SMTP  
(MX) I incorrectly thought they could all act in parallel.
If each  
domain should SMTP through a specific incoming server that's
fine (I  
assume that means the "not primary" incoming SMTP
servers would just  
be set to relay to the primary), but like Webmail I'd like
to have at  
least two servers able to take the incoming messages if
possible.

> I'm not saying it won't work, it's just not the way I'd
do it 

In general, I'd like to do things the way you recommend as
much as  
possible!   I thought
I was at least close to the preferred path  
by trying to follow the recommendations in the very helpful
responses  
to some support emails, and the diagram on the functionally
split  
cluster manual page, but thanks for your patience keeping me
on  
track 

Thanks!

Noah

>
> -- 
> When replying please quote original message to save
time and always  
> let
> us know what OS and version of SurgeMail you are
using.
>
>
************************************************************
********** 
> ******
> If you wish to communicate with other users and talk
about  
> SurgeMail and
> also get news about updates etc then join the SurgeMail
List.
> Send an email to surgemail-list-requestnetwinsite.com with only  
> "subscribe" in the message body
>
> FAQ - http
://www.netwinsite.com/surgemail/help/faq.htm
>
************************************************************
********** 
> ******
>