Oracle FAQ Your Portal to the Oracle Knowledge Grid

Home -> Community -> Usenet -> c.d.o.server -> Re: solaris 10 + 10gr2 + NOT a success story

Re: solaris 10 + 10gr2 + NOT a success story

From: joel garry <>
Date: 8 May 2007 15:11:57 -0700
Message-ID: <>

On May 8, 2:11 pm, "pmik" <> wrote:
> Hello everyone,
> We have recently installed 10gR2 RAC + ASM on 2 Sun servers running
> Solaris 10 OS. Since we started using the system (before applying
> patch we have been observing a rather strange behaviour.
> We start the CRS, nodeapps, asm and finally the database. crs_stat -t
> reports everything to be fine. Then, after a random period of time,
> one of the nodes' vip is lost, resulting to the eviction of the node.
> After applying the patch, the node is not evicted, the vip is
> reassigned to the other node and the instance keeps functioning
> properly. BUT, the vip continues to fail on random intervals.
> After searching through the OS's logs, in /var/adm/messages we get a
> message that the qfe (public IP nic) gets turned off and restored in
> a second at the specific moments that the vip service gets losts.
> After we experimented enough, we came to the discovery that if we
> keep pinging endlessly the public IPs of the the 2 nodes, the qfes of
> both the servers never get turned off and on. This way the RAC
> performs without any problems.
> Can someone of you explain this behavior? Is there something we are
> missing during or post installation? Is this a Solaris problem, an
> Oracle problem or a HW problem? We have almost eliminated the
> posibility for a NIC problem, since we switched the qfe nic for the
> ce nic (previously used for the interconnect) and we get the same
> behavior, this time for the ce card.
> Any help and/or guidance is extremely welcome since we seem to run
> out of options and directions. If anyone is interested, we will
> gladly provide any logs or elaborate on the matter.
> Thank you very much for your time and interest.
> Petros Mikos

I'll take a wild guess about something I know nothing about :-)

You are using containers, which are a virtualization of the NIC. This would be a problem, but Sun realizes that and allows you to define virtual nic control so data control structures can be maintained to the physical layer until no longer needed. So, you haven't quite correctly defined the control to work forever, it's timing out some time after being used. Pinging keeps the layer occupied. Use Sun support to work through this.



-- is bogus.
Costello: Tomorrow throws the ball. The batter bunts the ball. When he
bunts the ball, me being a good catcher, I wanna throw the guy out at
first base. So I pick up the ball and throw it to who?
Abbott: Now that's the first thing you've said right.
Costello: I don't even know what I'm talkin' about!
Received on Tue May 08 2007 - 17:11:57 CDT

Original text of this message