Re: [Vserver] clusters/farms/failovers?

From: Chuck <chuck_at_sbbsnet.net>
Date: Thu 21 Jun 2007 - 18:42:08 BST
Message-Id: <200706211342.08785.chuck@sbbsnet.net>

On Thursday 21 June 2007 03:14, Christian Affolter wrote:
> Hi Chuck!
>
> > I am in the idea stage for something I am not positive is entirely worth
the
> > effort and I am in totally uncharted territory.
> >
> > I am not sure what terminology is i am looking for here...
> Virtual servers, load balancing, fail over, high availability to name a
> few ;)
>
ok

> > what i am thinking i want to do is:
> >
> > have 2 hosts one a mirror of the other including all vserver guests etc.
the 2
> > machines are identical in every way.
> >
> > rather than have the mirror machine on 'standby' waiting for some fateful
day
> > it is needed, i would like both servers and all guests to be running
> > simultaneously. this would be accomplished by having everything running
> > unique private network ip addresses. this would allow adding
> > additional 'mirror' machines as necessary.
> >
> > the existing public ips from the production server we have running would
be
> > moved to some 'control' computer which would have a listing of the
private
> > ips that would serve what the public ip wants and would call on either one
as
> > needed. if one of the private ip servers doesnt respond (down) the control
> > computer would simply choose the working ip until the first one comes back
> > online.
> >
> > what do i need to do in the 'control' machine to accomplish this? is this
some
> > kind of configuration that already exists in the linux distro? we run
gentoo.
> > beowolf (whatever that is)?
> For the fail over or even load balancing in the future, have a look at
> the Linux Virtual Server project [1]. Probably the NAT approach [2]
> suits you best (as you want to use private IP addresses).
>

sounds good. will check that out.

> > is there another way of accomplishing what i wish to do? also how messy
will
> > keeping the mirror machines 'in sync' be? would i be better off having
all
> > machines but the controller share a common nfs mount for all the guests?
> Well it depends on what services you're going to run. If you want an
> active-passive setup (only one backend-servers is running at the same
> time) you "can" use NFS. However there are some services which are a
> little bit hairy to play nice with NFS mounts. Otherwise you can sync
> your data on the service level ~ replication (if the service provides
> the ability to do so). Last but not least DRBD block devices [3] is an
> option to keep machines in sync on the block device level.
>
> If you want to achieve an active-active setup consider to use some sort
> of SAN or DRBD running with a cluster file system like OCFS2 [4] or GFS
> [5] (to prevent locking and other issues).
>

will have to read up on this. i think this is what we need rather than trying
to sync 2 copies.

> > some of these virtual servers are very high volume usage so if all the
data
> > must route through the control computer, i am thinking that computer would
> > have to be a monster. or maybe have several control computers each
handling a
> > different class of service.
> I don't think they have to be a monster ;) However a fair amount of
> RAM would not be amiss.
>

after finding some information in other list archives, i see the control
doesn't need a lot of horsepower and that 1gb ram is good for approx 5
million simultaneous connections! so yeah. that machine can be downgraded
quite affordably.

> > sorry if i sound like i have no clue what i am talking about, but that is
the
> > truth :) , i only think i know what i want to accomplish.
> I'm sure there are other ways to accomplish this (Hardware load
> balancers etc.), nonetheless I hope I gave you some pointers in the
> right direction. Remember that this isn't the sort of setup which is up
> and running in 15min.
>

thank you.. i realize it will take a while. once i decide on a method and
begin study and implementing testing, i don't expect to be production ready
for a few months at least, then i would ease over less important services one
at a time while watching it carefully..

i did this with linux-vserver when i first discovered it.. although my
implementation time was considerably faster, that is due to the ease of setup
etc, i still took a good 4 months of testing before i set up a true
production machine. now i put vserver capability on everything including
workstations 'just in case i need a quick server for something' :) and in
fact my personal workstation has another 'workstation' vserver running kde
that i access via vnc for specialty work and i also allow some of our techs
to access so they can get graphical access to some servers available only on
our pvtnet which only my machines have access to.

thanks for the links too. saves a bit of time :)

>
> Good luck and regards
> Chris
>
>
> [1] http://www.linuxvirtualserver.org/
> [2] http://www.linuxvirtualserver.org/VS-NAT.html
> [3] http://www.drbd.org/
> [4] http://oss.oracle.com/projects/ocfs2/
> [5] http://www.redhat.com/software/rha/gfs/
>
> _______________________________________________
> Vserver mailing list
> Vserver@list.linux-vserver.org
> http://list.linux-vserver.org/mailman/listinfo/vserver
>

-- 
Chuck
"...and the hordes of M$*ft users descended upon me in their anger,
and asked 'Why do you not get the viruses or the BlueScreensOfDeath
or insecure system troubles and slowness or pay through the nose 
for an OS as *we* do?!!', and I answered...'I use Linux'. "
The Book of John, chapter 1, page 1, and end of book
_______________________________________________
Vserver mailing list
Vserver@list.linux-vserver.org
http://list.linux-vserver.org/mailman/listinfo/vserver
Received on Thu Jun 21 19:26:40 2007
[Next/Previous Months] [Main vserver Project Homepage] [Howto Subscribe/Unsubscribe] [Paul Sladen's vserver stuff]
Generated on Thu 21 Jun 2007 - 19:26:44 BST by hypermail 2.1.8