[gnhlug-announce] SLUG "problem solution" meeting Monday 7/14 7pm at UNH in Morse 301

Tom Fogal tfogal at io.iol.unh.edu
Thu Jul 10 11:27:22 EDT 2003


> 
> In a message dated: Thu, 10 Jul 2003 10:16:09 EDT
> Bob Bell said:
> 
> >On Thu, Jul 10, 2003 at 06:56:51AM -0400, Brian <lists at karas.net> wrote:
> >> I think many sites consider downtime to just be part of life for
> >> server moves, upgrades, etc.  Even "name brand" sites like ebay and
> >> Amazon have maintenance windows regularly where the site is pretty
> >> much offline an unavailable.
> >
> >    That's pretty stupid if they do.  Clustering technologies should let
> >you avoid this.
> 
> True, but there's really not a viable clustering solution on Linux/BSD
> which many of these sites are built on.  

i would beg to differ. openmosix can do clustering fairly easily. we run it
at home, but we havent tried service migration and such yet; its just home
use =)
i would -guess- that it works well enough, however. under BSD (well, at least
FreeBSD), there are available MPI implementations (what openmosix uses under
the hood), although ive never seen one that integrates with the kernel like
openmosix... i would looove to be proved wrong on that though.

a previous employment put me in charge of the 'backup' system - that is, starting
up external services when internal services went down. we didnt have a need for
second-only downtime, but the solution i designed could work in about five, minus
service startup time. there are a few a (free) projects that do service monitoring
and could be configured to monitor every second or so - although i would imagine
thats a bit much load to add. i suppose its a trade off with extraneous load
and response time, with such a solution.
i believe the system i designed utilized 'nefu' 
[http://rsug.itd.umich.edu/software/nefu/] for service monitoring. i remember it
working well enough but also vaguely remember seeing something that fit a little
better..

anyway, freshmeat searches should bring up such software.

> 
> MCL almost had a viable solution, but was very immature, and needed a 
> *lot* of work.  I don't know what's happened since they went away, or 
> if RH has enhanced Kimberlite enough to make it a viable option to 
> TrueClusters, etc.  I suppose it's possible :)

yea.. i have no clue about any of those. maybe i'm out of the loop.. =)

-tom

> -- 
> 
> Seeya,
> Paul
> --
> Key fingerprint = 1660 FECC 5D21 D286 F853  E808 BB07 9239 53F1 28EE
> 
> 	It may look like I'm just sitting here doing nothing,
>    but I'm really actively waiting for all my problems to go away.
> 
> 	 If you're not having fun, you're not doing it right!



More information about the gnhlug-discuss mailing list