[OpenAFS] Re: servers not establishing a quorum

lists@drewstud.com lists@drewstud.com
Tue, 6 Apr 2010 15:02:24 -0400 (EDT)

first off thank you again! I definitely appreciate everyone taking time to =
answer.=0AI removed the NetInfo file from the new openafs server we are try=
ing to add, and BAM quorum, the new afs server became the quorum, and you a=
re definitely correct about it having to do with what is and is not getting=
 registered (had to be since removing that file worked :) ) Each of our afs=
 servers so far is a fileserver and dbserver. =0A=0AAs far as sourcing ever=
ything on each server so that it comes from the vip via heartbeat, that is =
absolutely possible and we will probably go that route, since our servers a=
re both fileserver and dbservers and we will need some kind of vip to fail =
over. Good idea.=0A=0AThanks again for the help! I also finally understand =
how a quorum is elected much better after digging through the mailing list =
(found http://www.openafs.org/pipermail/openafs-devel/2001-January/005470.h=
tml).=0A=0A=0A-----Original Message-----=0AFrom: "Andrew Deason" <adeason@s=
inenomine.net>=0ASent: Tuesday, April 6, 2010 14:37=0ATo: openafs-info@open=
afs.org=0ASubject: [OpenAFS] Re: servers not establishing a quorum=0A=0AOn =
Tue, 6 Apr 2010 13:56:50 -0400 (EDT)=0Alists@drewstud.com wrote:=0A=0A> awe=
some.=0A> This may help as well:=0A=0A> we have afs "pairs" at each locatio=
n. We are syncing them with=0A> heartbeat/drbd.=0A=0ATrying to do that with=
 dbservers seems overkill, but okay. So you have a=0Ahot-spare thata starts=
 up bosserver when the other node goes down, I=0Aassume?=0A=0A> We have tri=
ed to get it to only "show" the one floating vip via=0A> NetInfo=0A=0AI hav=
en't been thinking about the cluster-HA AFS thing recently, but I'm=0Anot s=
ure how necessary that is. Fileservers will register what addresses=0Athey =
have on startup, so if the local IP is registered in the VLDB on=0Aone file=
server, and it goes down and the other server comes up, the old=0Alocal IP =
should go away. If/when clients re-read VLDB information, they=0Awon't get =
the IP for the downed fileserver.=0A=0A> VLLog=0A> Tue Apr  6 13:23:37 2010=
 ubik: primary address does not exist=0A> Tue Apr  6 13:23:37 2=
010 Using as my primary address=0A> Contents of NetInfo:=0A>= will work for fileservers, but I think for dbserv=
ers that's going=0Ato cause problems like the one you're seeing. When 10.13=
8.8.160 gets a=0Aping from, it doesn't know which site in the q=
uorum it=0Acorresponds to, since you told not to advertise the=
=0A172.20.1.26 address. Preferably for dbservers you would not specify=0Aan=
ything in that file.=0A=0AAlternatively, the easiest way for you to solve t=
his would probably be=0Ato just route outgoing packets such that they origi=
nate from=0A172.20.125.226 instead of (enabled with some heartb=
eat=0Ascript). Would that be possible?=0A=0A-- =0AAndrew Deason=0Aadeason@s=
S-info mailing list=0AOpenAFS-info@openafs.org=0Ahttps://lists.openafs.org/=