[techtalk] HA/Virtual Server solutions?

[Shane Landrum]

Hi all

So I started a new job with a company that does lots of work with open source
solutions, which is fun. One of the first tasks I've been put up to is
researching clustering technology for a client. The client has a cluster of
VA Linux webservers that are currently set up with round-robin DNS; they want a
more robust solution that'll route traffic away from dead webservers and do
load-balancing among the servers that are up. I've never touched clustering
technology, and this is a bit daunting to me. 

Can anyone recommend which software/packages/vendors I should look at
or stay away from? For example, I've been warned away from piranha because of
security issues in recent versions. Or, for that matter, can anyone tell
me what's behind the various cluster/HA products on the commercial market?
Most of them look like packaged versions of the LVS project's tools.

So far, I've looked at Understudy
(http://www.polyserve.com/prod_overview.html) and at packages that various
Linux vendors offer (Redhat, TurboLinux). I'm also looking at Ultramonkey,
which seems to be a decent noncommercial package of the Linux Virtual Server
project's tools.  How much of a pain is it to set up clustering with any of
these, and which tools offer the best feature set?

TIA,
srl
-- 
---
Shane R. Landrum + [EMAIL PROTECTED] + Software/Systems Engineer
-- www.ainnovations.com - Anansi Innovations --


___
techtalk mailing list
[EMAIL PROTECTED]
http://www.linux.org.uk/mailman/listinfo/techtalk

Re: [techtalk] Problem setting up dual-homed IBM NetFinity 1000

[Shane Landrum]

On Wed, 09 Aug 2000, C. M. Martin wrote:

> I have an IBM NetFinity 1000 running Caldera OpenLinux 2.4.  The system needs
> to be dual-homed, and the system seems to recognize both ethernet cards.  One
> is on eth0, IRQ11, and is working properly.  Network connectivity is just fine.
> The other is eth1, IRQ10, different I/O memory range from the first card.  The
> box can ping itself at that address, but can't see any other box on that
> network, and nothing on that net can see this box.  
> 
> I went into the network settings with COAS and found that the default gateway
> wasn't set on eth1.  I tried setting it, but the system would not save the new
> settings.  Sometimes it gives me a python error saying it cannot communicate
> with the network, sometimes not.  It reads:
> 
> Error when executing command ip-config change eth1
> the command produced the following output:
> SIOCADDRT: Network is unreachable
> eth1: initialization failed

It's probably a routing problem. Here's how to fix/troubleshoot it
interactively, though I don't know how to fix it with COAS.

Run "/sbin/route" and see what it gives you ("man route" if you need help
understanding what it says). Ifconfig just sets up the IP on the card; it
doesn't tell the kernel which interface to send packets through in order to get
to a certain net.  The kernel needs to know:

- that it can send packets to eth1 to get to itself
- that it can send packets to eth1 to get to the rest of the world.

IIRC, that means you need to do something like:

# /sbin/route add -host $IP_FOR_ETH1 eth1
# /sbin/route add -net default gw $YOUR_GATEWAY netmask $YOUR_NETMASK eth1

Let me know if this works; I haven't tested it, but I was playing with this
yesterday, so it's fairly fresh in my head.

srl
-- 
---
Shane R. Landrum + [EMAIL PROTECTED] + Software/Systems Engineer
-- www.ainnovations.com - Anansi Innovations --


___
techtalk mailing list
[EMAIL PROTECTED]
http://www.linux.org.uk/mailman/listinfo/techtalk