On Mon, 16 Jul 2007, Lars Marowsky-Bree wrote:
On 2007-07-14T20:52:43, Alan Robertson <[EMAIL PROTECTED]> wrote:
a random collection of mostly old cast-off computers. They range
between 300mhz and 2.4 ghz, and have disks ranging from
5200 to 7200 RPM with significantly different seek times.
Some support DMA, and I believe some of them don't, and the
amount of RAM also varies.
That cluster is possessed, but in a useful way.
I'd be interested in recreating those properties though, or making them
tunables in CTS.
If the most likely theory is that this is because of the timing
differences, that should be possible to simulate. On a virtualized
cluster, it's quite simple to limit CPU and memory assignments
accordingly. Or, within CTS, create an additional system load instead of
just running heartbeat.
given that the other 'interesting' cluster is a virtualized one there's alot of
potential in this.
what I'd suggest is make a _lot_ of virtual machines on one physical machine and
run them at different nice levels.
does anyone have some nice small HB configured virtual machines that can be used
rather then having to build them from scratch?
if not, let's talk before anyone build them.
<OFFTOPIC>
we need more documentation on the build process and how-to's. starting from
scratch to make a nice minimalistic install like you want for a VM is a good
time to document the minimum steps needed to make a useable cluster. (yes, I
know there are the two demos online)
I've been thinking of trying to make a series of demos that build on each other,
done with virtual machines the virtual images could be provided as well as
examples.
the type of things that I was thinking of are (not nessasarily in order)
minimal OS install
minimal firewall config (allow everything out)
make the example into a HA pair (adding state replication at some p
add ntp
add DHCP server for inside
add DNS server
add web proxy server (squid or equivalent)
add fileserver (samba and/or NFS), (including DRDB if it hasn't been done yet)
add a third machine to the cluster
distribute apps across the cluster
add IDS (snort)
etc,
this is the type of series that is never completely finished, but if it's done
as a series of building blocks it will hopefully be easy when someone asks 'how
can I do X' to point them at the right one, while keeping each of them simple.
the eventual goal that I'm thinking of is to create a general utility system for
a small office that would have every bell and whistle that a large company would
have. some of these things sound really complicated to some/most people (the
firewall configs for example), but if we can get people who are experts in the
areas to create sample best-practice configs
</OFFTOPIC>
I think those should help reduce the reliance on magic smoke ;-)
definantly a good thing.
Regards,
Lars
_______________________________________________
Linux-HA mailing list
[email protected]
http://lists.linux-ha.org/mailman/listinfo/linux-ha
See also: http://linux-ha.org/ReportingProblems
