Video Screencast Help

SEP LoadBalancing with Alteon ?

Created: 11 Jan 2008 • Updated: 22 May 2010 | 5 comments

hello,

i'm looking for the right configuration of my servers in order to do this :

- loadbalancing of web ( http or https) requests done by the sep clients to 2 sep servers, done by Alteon.
- replication of SQL server in order to avoid loss of database if one fails.

I think i should install like this :

First server : SEP with remote sql server that point to the "third server"
Second server : SEP with remote sql that point to the "third" server too

Third server : SQL server
fourth server : SEP + SQL that is a "remote site for replication" ?

am i right ?
documentation does not talk about it , nor the kb ...

thank you in advance,

regards.

Comments 5 CommentsJump to latest comment

lawman 2's picture

Yes your config you’d definitely accomplish what you want.  Just few points for your consideration.  Symantec has basically 4 recommended architectural designs Centralized, Distributed, Log replication, and High Availability.  In the HA scenario  there is really no need for a load balancer since it can be accomplished by simple DNS round robin but if you have one than why not use it. As for database HA, a clustered database would be my first choice and then replication for DR/loss of site.  Veritas cluster server is a great way to cluster the database

Here is some info on the recommendations I found.

 

 

 

 

 

 

 

 

 

xafiquet's picture

wow! nices slides !

in fact, our loadbalancer is hosting a "dns vip", so that , users does not know the real ip of our managements servers.

If i want to use the SEP replication, how it works if my "master sql server failed" ? i mean, my SEP servers know that they have to update/read from the "replication sql server" ? or it is a "manual" procedure that i have to do when i see my "sql master server" down ?

thx a lot for your answer, appreciate it.


lawman 2's picture
Ok,  the fact that you load balancer is hosting a DNS vip is fine.  The clients don't need to know the real address of the management server.  They just need to connect to one at the address or name specified in the Management Server list specified in the policy applied.
 
So let look at replication.  In SEPM replication takes place between 2 SEPM servers not at the database.  So if at the primary location the SQL server goes down the SEPM managers using that database can not function as they really don't know about the other SQL server you were replacing to since all they know about is the other SEPM server.  This is why it is recommended for a HA environment you cluster your SQL servers.  If you are using Veritas Cluster Server for you SQL database it will handle the database failover from one SQL server to another.  And just like with the clients and the loadbalancer the SEPM does not really know the true IP of the SQL database.  It only knows about a virtual IP controlled by Veritas cluster server.  So that IP will fail the SQL database over.  If you lose the whole site SEPM and Database from say a building disaster the clients can be configured to go look for the replicated info in the DR site.
 
I hope this helps. 
xafiquet's picture

hello,

thank you for your answer.

is is possible to get this "clustered db" using the native  sql server clustering ? what are the other way to make the DB in high availability ? i mean, veritas is a big piece to accomplish i think!

regards

lawman 2's picture
Clustering the DB is more of a preference.  It is recommended to use MCS or VCS.  I prefer VCS as I am more familiar with it.  I believe MS requires Server 2003 Enterprise to cluster but that might not be the case any more.  VCS can cluster 2003 standard edition and the offset the cost.  Plus with VCS there is no quorum disk witch is normally a single point of failure with MCS.  You might want to talk to your DBA and see how they do HA.