|
Hello,
On Tue, 2 Jan 2001, ratz wrote:
> Hi Julian,
>
> > - resource information for Linux, FreeBSD, Solaris, HP/UX, WinNT 4,
> > SCO, Unixware. I still don't have load parameters for Win2K.
>
> What about SNMP? Do they have now some kind of intelligent snmpd?
I don't like the SNMP protocols :) I prefer to split the
source to independent modules. Then we can build different daemons and
tools for the clusters. I hope the code will speak better in the next
weeks.
> > - redundancy: primary and backup directors, different role for different
> > clusters
>
> I've been at the CCC Congress in Berlin lately and I met one of the
> netfilter core developers and he actually is already working on the
> problem for kernel 2.4 and iptables. I'll contact him and ask about
> the status of his work. As far as I spoke to him he's also trying to
> keep the iptables structures in synchronisation with a backup node
> and I reckon this will be pretty much the same code for LVS.
Oh, no. I don't talk for active-active setups. I don't have
good ideas for active replication. Only bad :) I'm talking for setups
where one/many of the real servers can be a backup server and to take
the control when the main director fails. It happens sometimes but
not very often, here is one of my directors:
uptime:
11:33am up 98 days, 7:41, 1 user, load average: 5.41, 3.86, 3.43
See, load is 5.41 - busy with healthchecks and monitoring. This is the
reason I prefer to avoid these healthechecks in the director.
Kernel:
Linux vs 2.2.18pre10 #1 SMP Wed Sep 27 03:49:24 PST 2000 i686 unknown
LVS:
IP Virtual Server version 0.9.16 (size=16384)
You see, can you believe I need active-active replication :) The only
problems come from the f***ing databases :( For me, the main advantage
would be in a good load balancing software, with user-defined rules and
behavior.
> Looking forward,
> Roberto Nibali, ratz
Regards
--
Julian Anastasov <ja@xxxxxx>
|
