It is important, but it DO NOT WORK. That's the point. It DO NOT MIGRATE in 50% scenarios. I tried to tropubleshout it few times, but found much simpler to give up and forget about this thing at all (why? see below).
Problem is that VIP do not provide correct service failover. When new node comes, system FIRST switch VIP (so you LOST all your services at once), and then (in a FEW minutes) starts new services on these node. So, it do not provide real failover (which should be - start services FIRST and THEN switch VIP; or switch VIP first and then shutdown services). We discussed at on www.sql.ru forum some time ago, and conclusion was simple - VIP is not useful thing for today.
So, if I use VIP for uninterruptebla service, I wil have a few minutes interruption every time when service switchover between nodes. So, I betyter use Loadbalancer (with 5 second monitoring interval) which failover me to REALLY WORKING node without any interruptions (and keeps opened sessions where they are already).
(SR - what SR? They had these problem in most of their RAC installations. And it do not influence any applications. It is minor bug, after all they have with RAC and clusters).
PS Thanks for a very useful information about crs usage.
PPS. There is one problem with SR's. If it is critical bug or production system, Ialways popen TAR or SR. But if I see something which looks like a bug, and I want to inform Oracle about it (but do not ask to assign engineer except if it is problems for others as well and Oracle want to troubleshout it with me), how can I do it?
-- -- Original Message -- -- From: <CLEMENS.BLEILE@(protected)> To: "Alexei_Roudnev" <Alexei_Roudnev@(protected)>; "Silviu Marin-Caea" <silviu_marin-caea@(protected)>; <suse-oracle@(protected)> Cc: <clemens.bleile@(protected)> Sent: Wednesday, February 08, 2006 2:29 AM Subject: AW: Re: [suse-oracle] When should a VIP address move to other node?
> > Hi, > > the VIP is a very important feature and you *must not* forget about the VIP. > > The only reason why we have a VIP is to avoid TCP/IP-timeouts. And this is crucial in a high availability environment. It's not acceptable for a client to wait minutes for the timeout. > > If the VIP is not moving over during a node shutdown (it actually should) then contact Oracle Support and open a SR. > > The main thing, which was missing for me in the context of VIPs was the feature of gracefully switching a VIP over to another node in case of e.g. planned maintenance. Therefore Bug 4623630 was opened to enhance srvctl. A new option "-r" (relocate) is added to "srvctl stop nodeapps" command that automatically relocates all active VIPs on the node to another node or stop all active VIPs when this is the last node. > > If you want to stop crs and move a VIP to another node you may follow the following steps as an "unsupported" workaround. > > 1. srvctl stop instance -i < instance> -d < database> > for all databases that have instance configured on the < node> > 2. srvctl stop asm -n < node> > 3. srvctl stop nodeapps -n < node> > 4. crs_start ora.< node>.vip -f > 5. crs_relocate ora.< node>.vip -f > 6. /etc/init.d/init.crs stop > > Using "crs_"-commands is only supported for resources which do NOT start with "ora", ie it is supported only for your own applications, which you want to make high available (you always should use srvctl for resources, which start with "ora"). You therefore better ask for a backport of enhancement request 4623630. > > Ciao > > Clemens > > >
-- To unsubscribe, email: suse-oracle-unsubscribe@(protected) For additional commands, email: suse-oracle-help@(protected) Please see http://www.suse.com/oracle/ before posting
