Short answer: Yes, it can be done. Thanks to everyone who replied (I never mention names since some people want to remain anonymous, but I appreciate all replies). Since there have been many questions lately about IP Multipathing, here are links to some docs: http://www.sun.com/blueprints/0201/Multipathing.pdf http://www.samag.com/documents/s=1441/sam0111i/0111i.htm http://docs.sun.com/ab2/coll.780.2/IPNETMPADMIN/@Ab2TocView/315?Ab2Lang=C&Ab2Enc=iso-8859-1 http://www.eng.auburn.edu/~doug/howtos/multipathing.html I asked: > I know that in Solaris8, you can use IP Multipathing (in.mpathd) to > provide failover across a pair of ethernet interfaces. > > I also know that with 2 physical interfaces, you will need a > total of 4 IPs > (two real, two dummy). Only one is the "real" address that > talks to the > outside world, and would accept incoming traffic (so my DNS > would point to > this address). > > For this example, presume that the "real", visible IP is > 192.168.10.10, on > hme0. If hme0 fails, hme1 will take over with the same IP address. > > Now for my question: in olden times, it was common to bind > more than one IP > to a single ethernet interface, for purposes of Apache virtual hosts. > Assume I still want to do virtual hosts that way. > Is it possible in the multipathing example above, to have > more than one public > IP bound to the active interface? > > To wit, can I have the machine such that: > 192.168.10.10, 11 and 12 are on hme0 > if hme0 fails, all 3 IPs get moved to hme1 > > I've reviewed all the docs from Sun and cannot find a > conclusive answer. Short answer: Yes, it can be done. Detailed answers: yes. all non-deprecated IPs move in a failure. -- I have not tired this, but I believe that you would just have to add more addif statements for the interface hme0 to give you the individual virtual interfaces that you are looking for. I would try this, but it would affect our production systems.. so I can't. If this works, please let me know so that I can use it! Here is what I would think the hostname.hme0 and hostname.hme1 would look like. /etc/hostname.hme0 cador-hme0 netmask + broadcast + \ group production deprecated -failover up \ addif cador0 netmask + broadcast \ addif cador1 netmask + broadcast \ addif cador2 netmask + broadcast + failover up /etc/hostname.hme1 cador-hme1 netmask + broadcast + \ group production deprecated -failover up \ addif cador3 netmask + broadcast \ addif cador4 netmask + broadcast \ addif cador5 netmask + broadcast + failover up /etc/hosts 192.168.10.10 cador0 192.168.10.11 cador1 192.168.10.12 cador2 192.168.10.13 cador3 192.168.10.14 cador4 192.168.10.15 cador5 192.168.10.20 cador-hme0 192.168.10.21 cador-hme1 -- Yes. Just mark that interface with the 'failover' flag and it will be part of the IPMP process. ie: $ cat hostname.hme0 <- dummy bigbad1-test1 netmask + broadcast + group bigbad1 deprecated -failover up $ cat hostname.hme0:1 <- real bigbad1 netmask + broadcast + failover up $ cat hostname.hme0:2 <- real bigbad1-virt1 netmask + broadcast + failover up _______________________________________________ sunmanagers mailing list sunmanagers@sunmanagers.org http://www.sunmanagers.org/mailman/listinfo/sunmanagersReceived on Thu Dec 20 11:53:33 2001
This archive was generated by hypermail 2.1.8 : Thu Mar 03 2016 - 06:42:31 EST