|
On Thu, 14 Oct 1999, Zalman Stern wrote:
> Yes, this is exactly what I want to do. Except that I'd like to do it such
> that all the servers (well only two) are really running on a single
> machine.
I don't think LVS allows you to redirect within amachine like that - but
that's not something I've needed to do - I'm sure other people here more
knowledgeable in these thingscan put me right on that department. (*Maybe*
some IP chains config would help here?) I dunno.
> Outside of the above setup, I'm also looking at using LVS for
> high-availability using multiple real-servers.
I've been to busy dealing with other stuff really to look back in this
list, but one thing I can say is that depending on what sort of load
you're likely to put through the system LVS does seem to be really stable,
and reliable enough to do the job it says it does :-)
Currently we're testing the LVS as a means of load balancing our national
caching infrastructure, and we're now half way through the primary test
phase. Since this is going extremely well so far, assuming the tests
continue to go well, we'll probably have a pilot service in place in a few
weeks time.
What do I mean by going well ? Well in terms of requests/data shipped per
day, one LVS master is balancing 3 servers off the back of it, to the tune
of 5.25 million http proxy requests per day, or put another way - about
40Gb per day. This is well within the tolerances of the backend servers.
Response times across the test cluster have been very consisent, and load
balancing has been close to what could be considered optimum.
During testing I've so far simulated real-server death a few times (with a
suitable monitoring system in place) and the system has failed over much
more gracefully than it does at the moment.
There's still a way for us to go before we move this into full production
service and put the rest of our traffic through the system including quite
a few tests but if you want to load balance about 5 million http requests
per day or 40Gb of TCP traffic, LVS is probably the thing you want. (The
amount of traffic we need to balance is an order of magnitude greater than
this hence the all test phases!)
If people would find it useful I could post summaries (or pointers to
summaries) of the test results here as well... (Whilst most people here
probably have confidence in the system, most sysadmins probably do what I
did when looking at the stuff - looked through the mailling list archives
looking for "success reports", and lets face it the more there are the
more likely management will let things like this through :-)
Michael.
--
National & Local Web Cache Support R: G117
Manchester Computing T: 0161 275 7195
University of Manchester F: 0161 275 6040
Manchester UK M13 9PL M: Michael.Sparks@xxxxxxxxxxxxxxx
----------------------------------------------------------------------
LinuxVirtualServer.org mailing list - lvs-users@xxxxxxxxxxxxxxxxxxxxxx
To unsubscribe, e-mail: lvs-users-unsubscribe@xxxxxxxxxxxxxxxxxxxxxx
For additional commands, e-mail: lvs-users-help@xxxxxxxxxxxxxxxxxxxxxx
|
