LVS
lvs-users
Google
 
Web LinuxVirtualServer.org

Re: Network Block Device w/ RAID for Content Replication (Was: Re: load

To: Joseph Mack <mack@xxxxxxxxxxx>
Subject: Re: Network Block Device w/ RAID for Content Replication (Was: Re: load balance web site)
Cc: Atif Ghaffar <atif@xxxxxxxxxxxx>, Andi Hechtbauer <anti@xxxxxxx>, tc lewis <tcl@xxxxxxxxx>, Joseph Mack <mack.joseph@xxxxxxx>, <pu@xxxxxxxxxx>, <lvs-users@xxxxxxxxxxxxxxxxxxxxxx>
From: Atif Ghaffar <atif@xxxxxxxxxxxx>
Date: Sat, 03 Jun 2000 16:01:55 +0200
Quoting Joseph Mack <mack@xxxxxxxxxxx>:

nbd looks like a normal file system.
I am not using this device to writes at all.

Infact the applications know that there is a
ro and rw mail, ldap and mysql server.

ro is localhost 
and rw is the ip of the back end server.


The writes are performed only on the backend servers.
The front end servers only read from those disks.

Example (mysql is running on all frontend and the backend node)
This way instead of putting heavy load on one mysql server we use all the front
end servers.


LDAP has builtin replication but nbd is giving the same thing.


Ofcoarse the content modification (uploads etc) are done by the development team
on the main cvs server.

> does nbd look like nfs (or a regular filesystem) in that when 2
> processes
> attempt to write to a file, one of them will see a write lock (eg if
> the user is logged on twice to different front end servers, and running
> 2
> copies of their mail reader, will the 2nd mailreader see that the mail
> spool file is write locked)?
> 
> do you have 2 mysqld's running? How do you do the write locking?
> 
> Joe
> 
> --
> Joseph Mack mack@xxxxxxxxxxx
> 
> 



--
    Perfection is our goal, excellence will be tolerated. -- J. Yahl


<Prev in Thread] Current Thread [Next in Thread>