[ Date Index ] [ Thread Index ] [ <= Previous by date / thread ] [ Next by date / thread => ]
Sam Grabham wrote: > > I don't really want to risk rebooting , in case it doesn't come back up > again. > > I would like to if possible to cluster this box. Urm you can't get the "ls" command to work reliably on it, and you are thinking of clustering it. I think cart and horses need rearranging here. What else does it do beside routing, as it looks over specced for a router unless the leased line is something fantastic. Obviously a lot of clustering issues are determined by acceptable downtime. But I'm skeptical of the benefits for most folk. Typically clustering is quite complex, and risks introducing new failure modes, so you need to be absolutely sure of the design and hardware requirements before trying to do this sort of thing. Otherwise you'll end up with lower, not higher reliability. Heck I have Linux boxes acting as routers for years on end, everything sits in RAM, the disks don't spin, and they usually fail when one of the fans wears out. In one case the disk did fail, but nothing noticed till the next planned service point, when shutdown caused a panic. If it is only routing traffic, than one can set up routers in parallel, such that if one fails the other routes its packets with no clustering technology - just routing algorithms. But again it all depends on details of the network configuration. Failing that you could fail the IP address between the two, and provide routing using the virtual IP address. So the question should be "what does it do?" "What is acceptable down time (planned and unplanned)" and work from there. -- The Mailing List for the Devon & Cornwall LUG http://mailman.dclug.org.uk/listinfo/list FAQ: http://www.dcglug.org.uk/linux_adm/list-faq.html