fault-tolerance on linux servers?

fault-tolerance on linux servers?

Post by Jeff Howar » Wed, 12 May 1999 04:00:00



Hi all,

I need a bit of advice on setting up a couple of linux servers to be, to use
a buzz word, fault tolerant.  At present I have a couple of big NT servers
with a (IMHO) fairly junky bunch of software that is supposed to allow one
of the machines to take over the functionality of the other machine should
one of them fail.  The problem being: 1:  The servers are NT servers.  'Nuff
said.  2: The fault-tolerance software causes more problems than it solves.
(Random crashes, etc) 3: The services running on the NT servers are
mission-critical.
The servers are currently configured with 2 NICs apeice. The first one on
each machine goes to an ethernet switch, and the second nic is connected via
a crossover cable to the second nic on the other server.  Both servers are
running different services.
Finally, to the heart of my question, how can I set up 2 (or more, perhaps?)
linux servers in such a way as to allow one of them to detect a failure in
the other and to assume the duties that the other server was handling?
I've tried to look into clustering unix servers but all I keep running into
is Beowulf clusters.  Is that what I need to study up on?
Any help/advice on this will be greatly appreciated. (Plus you get the added
benefit, and the warm, fuzzy feeling that comes with replacing 2 more NT
servers with linux ;-)

Thanks in advance,
Jeff Howard

to email me remove my opinion of spam

 
 
 

fault-tolerance on linux servers?

Post by Curt » Wed, 12 May 1999 04:00:00


Take a look at
http://metalab.unc.edu/pub/Linux/ALPHA/linux-ha/High-Availability-HOW...
has some links for commercial stuff as well.

Linux High Availability project http://www.henge.com/~alanr/ha/

I've not played with any of it yet, but am very interested in hearing more.


> Hi all,

> I need a bit of advice on setting up a couple of linux servers to be, to
use
> a buzz word, fault tolerant.  At present I have a couple of big NT servers
> with a (IMHO) fairly junky bunch of software that is supposed to allow one
> of the machines to take over the functionality of the other machine should
> one of them fail.  The problem being: 1:  The servers are NT servers.
'Nuff
> said.  2: The fault-tolerance software causes more problems than it
solves.
> (Random crashes, etc) 3: The services running on the NT servers are
> mission-critical.
> The servers are currently configured with 2 NICs apeice. The first one on
> each machine goes to an ethernet switch, and the second nic is connected
via
> a crossover cable to the second nic on the other server.  Both servers are
> running different services.
> Finally, to the heart of my question, how can I set up 2 (or more,
perhaps?)
> linux servers in such a way as to allow one of them to detect a failure in
> the other and to assume the duties that the other server was handling?
> I've tried to look into clustering unix servers but all I keep running
into
> is Beowulf clusters.  Is that what I need to study up on?
> Any help/advice on this will be greatly appreciated. (Plus you get the
added
> benefit, and the warm, fuzzy feeling that comes with replacing 2 more NT
> servers with linux ;-)

> Thanks in advance,
> Jeff Howard

> to email me remove my opinion of spam


 
 
 

fault-tolerance on linux servers?

Post by Erik Norvell » Wed, 12 May 1999 04:00:00


Check out the LInux high-availability site.  It's located at
http://www.henge.com/~alanr/ha/.  It's still in its infancy, but there are a
lot of links to "fault tolerance" open source projects that have actually made
a lot of progress.  Especially check out the Linux Virtual Server project which
has a link from the High Availability site.

-Erik Norvelle


>Hi all,

>I need a bit of advice on setting up a couple of linux servers to be, to use
>a buzz word, fault tolerant.  At present I have a couple of big NT servers
>with a (IMHO) fairly junky bunch of software that is supposed to allow one
>of the machines to take over the functionality of the other machine should
>one of them fail.  The problem being: 1:  The servers are NT servers.  'Nuff
>said.  2: The fault-tolerance software causes more problems than it solves.
>(Random crashes, etc) 3: The services running on the NT servers are
>mission-critical.
>The servers are currently configured with 2 NICs apeice. The first one on
>each machine goes to an ethernet switch, and the second nic is connected via
>a crossover cable to the second nic on the other server.  Both servers are
>running different services.
>Finally, to the heart of my question, how can I set up 2 (or more, perhaps?)
>linux servers in such a way as to allow one of them to detect a failure in
>the other and to assume the duties that the other server was handling?
>I've tried to look into clustering unix servers but all I keep running into
>is Beowulf clusters.  Is that what I need to study up on?
>Any help/advice on this will be greatly appreciated. (Plus you get the added
>benefit, and the warm, fuzzy feeling that comes with replacing 2 more NT
>servers with linux ;-)

>Thanks in advance,
>Jeff Howard

>to email me remove my opinion of spam

 
 
 

fault-tolerance on linux servers?

Post by Raymonds Doetje » Wed, 12 May 1999 04:00:00


I've already seen that some-one gave the LINUX-HA project idea to you.
There is a piece of software called VINCA that does the same for NT and Novell.
I must say it is a pretty good failover software. Just like the LINUX-HA.

But the best fail-over software is still NOVELL Netwares 4.11 SFTIII no
downtime at all when a server goes down scheduled/or non-scheduled. If your
missiom critical server is fileserving than this is the best solution for you.
But seeing you run NT it is probably some applicationserver.

Raymond


> Hi all,

> I need a bit of advice on setting up a couple of linux servers to be, to use
> a buzz word, fault tolerant.  At present I have a couple of big NT servers
> with a (IMHO) fairly junky bunch of software that is supposed to allow one
> of the machines to take over the functionality of the other machine should
> one of them fail.  The problem being: 1:  The servers are NT servers.  'Nuff
> said.  2: The fault-tolerance software causes more problems than it solves.
> (Random crashes, etc) 3: The services running on the NT servers are
> mission-critical.
> The servers are currently configured with 2 NICs apeice. The first one on
> each machine goes to an ethernet switch, and the second nic is connected via
> a crossover cable to the second nic on the other server.  Both servers are
> running different services.
> Finally, to the heart of my question, how can I set up 2 (or more, perhaps?)
> linux servers in such a way as to allow one of them to detect a failure in
> the other and to assume the duties that the other server was handling?
> I've tried to look into clustering unix servers but all I keep running into
> is Beowulf clusters.  Is that what I need to study up on?
> Any help/advice on this will be greatly appreciated. (Plus you get the added
> benefit, and the warm, fuzzy feeling that comes with replacing 2 more NT
> servers with linux ;-)

> Thanks in advance,
> Jeff Howard

> to email me remove my opinion of spam

 
 
 

fault-tolerance on linux servers?

Post by Claud » Thu, 13 May 1999 04:00:00


Quote:> (Random crashes, etc) 3: The services running on the NT servers are
> mission-critical.

Can u get those same applications (services) for linux ?

Claude

 
 
 

fault-tolerance on linux servers?

Post by Jeff Howar » Thu, 13 May 1999 04:00:00


Yes. They're all standard intranet services (DNS, HTTP, FTP, RADIUS,
etc...).

>> (Random crashes, etc) 3: The services running on the NT servers are
>> mission-critical.

>Can u get those same applications (services) for linux ?

>Claude