LVS
lvs-users
Google
 
Web LinuxVirtualServer.org

Re: Where to get started

To: "LinuxVirtualServer.org users mailing list." <lvs-users@xxxxxxxxxxxxxxxxxxxxxx>
Subject: Re: Where to get started
From: "Karl Kopper" <karl@xxxxxxxxxxxxxxx>
Date: Mon, 9 Feb 2004 10:08:12 -0800
> Hello
> I am very familiar with the linux env. But i now need to go to the next
step
> and have some questions just to get me going in the correct direction
>
> I need to build 2 (initially ) redhat boxes. (probably  RHS - ES ) I
really
> like RH9 but need the up2date stuff.

Yum might help here. Though I don't know which direction they are going
after 9.0.

> There is also "Red Hat Cluster Suite", but being a small startup, we are
> trying to keep costs down as much as possible.
>
> With the boxes i am building , they really need to be clustered. I
> understand the concept of clustering, but i have a couple of gaps that i
do
> not understand. (BTW - is there a clustering for dummies? )
>

The best documentation right now is available on-line. Probably the LVS FAQ.
Hopefully something good will be in print soon. :)

> The biggest question is how do the clusters share data? Is this done via
> hardware or software. In the LVS I see lots of stuff about setup for load
> balancing, which is part of what i need.
>

To load balance requests (connections) you'll need to share data and
arbitrate access to the data with a locking mechanism if the data is
writeable. Unless you use something like the ZODB with Zope with its
multiversion concurrency control (MVCC) system. The Zope web site, by the
way, says it is a clustered LVS web site.

Local data can be stored on each node and someting like GFS (now owned by
Red Hat, I believe) can arbitrate write access. However, cluster nodes make
good NFS clients (NFS is very stable on Linux--at least I can say this about
the 2.4 kernel) and there are lots of good NAS solutions out there. If costs
are a big concern you can build your own NFS server, but you might be up
against a performance wall.

> Is LVS what i need?  ( my app is a web based app, that reads and writes to
> datafiles on the webserver)
>

If you want high availability you'll want LVS and keepalived or Heartbeat
(Heartbeat+STONITH+ldirectord) for high availability.

> Can anyone point in the direction to get started?
>

The Heartbeat and LVS HOWTOs are a good place to start IMO. Also, you might
want to take a look at OSCAR--one of the goals of OSCAR is to help people
get started with clusters and it has LVS + Heartbeat bundled. For easier
admin there is also the webmin project if you want a GUI, and for monitoring
check out Mon from kernel.org and/or the Ganglia project.

--Karl

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