Re: Shutdown nodeapps does not bring down VIP and Listener nodeapps

On 9 1 , 3 47 , FatP <thats..._at_gmail.com> wrote:
> On Sep 1, 12:26 am, Marc Bruinsma <marc.bruin..._at_chello.nl> wrote:
>
>
>
> > FatP wrote:
> > > I am learning 10gR2 RAC on ubuntu and have gone through the
> > > installation of Clusterware successfully (no DB yet). But I am facing
> > > some issues:
>
> > > 1. Shutdown nodeapps does not bring down VIP
> > > When I use the following command to shutdown nodeapps
>
> > > srvctl shutdown nodeapps -n [nodename]
>
> > > all the Oracle processes are died. However, eth0:1, which is used by
> > > VIP is not brought down. Then srvctl refuses to start nodeapps because
> > > the IP address is occupied. As a result, I need to use root to bring
> > > down eth0 then assign the default route to the gateway again. Is this
> > > the expected behaviour?
>
> > > Actually, the result is the same if I shutdown CRS using /etc/rc.d/
> > > rc3.d/init.crs.
>
> > > 2. Cannot start listener with nodeapps
> > > When I start nodeapps, it complained that LISTENER_[NODENAME] cannot
> > > be started. Am I missing something Is this a database listener (so
> > > that after I install & config db, the error will disappear) or
> > > nodeapps has another listener?
>
> > > Environment:
> > > Oracle 10g R2 x86
> > > Ubuntu 6.10 (I know this thing is not supported. But I am just
> > > learning... and other things look good anyway)
>
> > Which version are you using exactly? Are all the required packages on the
> > nodes (Oracle is very picky about this).
>
> > When you installed clusterware, did the vipca script run succesfully? Almost
> > every cluster install I did, I had to rerun it manually from the last node
> > in order to create the vip's succesfully.
>
> > It is absolutely crucial that all the IP adresses and hostnames are
> > configured correctly in /etc/hosts on all the nodes before you run the
> > install, otherwise you won't get it to function properly.
>
> > Or did you by any chance configure a second IP on eth0 manually by creating
> > an ifcfg-eth0:1? I've seen that happen once before.
>
> > What does srvctl config nodeapps -n <node> show you, and is it what you
> > expect?
>
> > BTW: an better way to shut down CRS is by crsctl stop crs.
>
> > That's it for now, dinner is being served...:-0
>
> Yes it is srvctl stop. Stupid typo.
>
> I searched through all the log files in ORA_CRS_HOME, and concluded
> the problem occurs when running racgvip. In fact, the same error
> messages appeared when I ran racgvip in "test mode"...
>
> root:$ORA_CRS_HOME/bin# export _USR_ORA_VIP=192.168.128.129
> root:$ORA_CRS_HOME/bin# export _USR_ORA_NETMASK=255.255.255.0
> root:$ORA_CRS_HOME/bin# export _USR_ORA_IF=eth0
> root:$ORA_CRS_HOME/bin# export _CAA_NAME=ubuntu
> root:$ORA_CRS_HOME/bin# ./racgvip start
> root:$ORA_CRS_HOME/bin# ./racgvip check
> IP:192.168.128.129 is already up in the network (host=ubuntu)
>
> (This showed VIP was brought up correctly. ifconfig confirmed that
> eth0:1 has IP address 192.168.128.129)
>
> root:$ORA_CRS_HOME/bin# ./racgvip stop
> bounce eth0 (host=ubuntu)
> SIOCSIFADDR: File exists
> SIOCSIFFLAGS: Cannot assign requested address
> SIOCSIFNETMASK: Cannot assign requested address
> SIOCSIFBRDADDR: Cannot assign requested address
> SIOCSIFFLAGS: Cannot assign requested address
> SIOCSIFFLAGS: Cannot assign requested address
>
> ifconfig showed that eth0:1 was still there. So racgvip could not
> bring down the VIP logical interface.
>
> Further tracing showed that the actual commands executed by racgvip
> were:
>
> /sbin/ifconfig eth0:2 192.168.128.129 netmask 255.255.255.0 broadcast
> 192.168.128.255 up
> /sbin/ifconfig eth0:2 del 192.168.128.129
>
> Changing eth0:2 to eth0:1 brings up the VIP interface.
>
> It is 10g R2 (10.0.2.0.1). Since it is not a supported OS, it is hard
> to say whether all the requested packages are installed. But I believe
> what I have is OK after reading the numerous "Installing Oracle on
> Debian" guides.
>
> vipca did not run correctly during install. I need to manually run
> oifcfg to config the interfaces, then things looks ok...
>
> "srvctl config nodeapps -n <node>" shows nothing. Any suggestion?

I found out what the problem is. ubuntu ships mawk, which does not support character class [[:alnum:]]. This makes getifbyip unable to get eth0:1 as the correct interface and eventually racgvip attempts to bring down eth0:2.

Solution:

In racgvip, around line 215: change the line gf_retif=`$IFCONFIG $2 | $AWK "BEGIN { name =\"\" } { if (/ ^[[:alnum:]]/) name=\\$1; if (/inet addr:$__LOCAL_IP /) { print name } }"`

to

gf_retif=`$IFCONFIG $2 | $AWK "BEGIN { name =\"\" } { if (/^[a-zA- Z0-9]/) name=\\$1; if (/inet addr:$__LOCAL_IP /) { print name } }"` Received on Mon Sep 03 2007 - 01:53:44 CDT