Re: 11gR2 RAC Network Failure - Timeout
Date: Mon, 6 Feb 2012 13:08:51 -0600
You may want to use -p flag to see those attributes of a resource. Check_interval is set to 1 second and that's the interval between two checks for that VIP resource. VIP resource also has a check_timeout attribute set to 30 seconds. So, it might be approximately few seconds after 30 seconds for the VIP failures to be detected. It is possible that clusterware is trying to tear down the dependent resource causing 10 more seconds of delay.
For network resource, script_timeout is set to 60 seconds, if the thread executing the script does not complete in 60 seconds for a resource, that resource is considered not responding.
Of course, you shouldn't modify these attributes as per the documentation.
$ crsctl stat res ora.some.vip -p
DESCRIPTION=Oracle VIP resource
START_DEPENDENCIES=hard(ora.net2.network) pullup(ora.net2.network) START_TIMEOUT=0 STATE_CHANGE_TEMPLATE=
$ crsctl stat res ora.net1.network -p
Ora!nternals - http://www.orainternals.com - Specialists in Performance, RAC and EBS
OakTable member http://www.oaktable.com and Oracle ACE Director
On Mon, Feb 6, 2012 at 12:28 PM, D'Hooge Freek <Freek.DHooge_at_uptime.be>wrote:
> I think that sets only the interconnect timeout, not the vip timeout.
> On linux the vip connectivity is checked by using ethtool (or mii-tool on
> older linux versions) to verify the status of the network port. If ethtool
> reports that the link is down a ping is done to the public gateway. If the
> ping test fails two times, the failover is initiated.
> Looking at the script, I think the timeout of the ping is set to 1 second
> (-w 1 -c 1).
> The script responsible for the vip test is $ORACLE_HOME/bin/racgvip (where
> $ORACLE_HOME is the GI home, not the db home).
> On linux this is just a shell script, so you can modify it to change the
> number of retries.
> But I can't directly find where the sleep time between two vip checks is
> Freek D'Hooge
> Oracle Database Administrator
> email: freek.dhooge_at_uptime.be
> tel +32(0)3 451 23 82
> disclaimer: www.uptime.be/disclaimer
> -----Original Message-----
> From: oracle-l-bounce_at_freelists.org [mailto:oracle-l-bounce_at_freelists.org]
> On Behalf Of Phillip Jones
> Sent: maandag 6 februari 2012 16:13
> To: puravc_at_gmail.com
> Cc: oracle-l_at_freelists.org
> Subject: Re: 11gR2 RAC Network Failure - Timeout
> To view the timeout:
> crsctl get css misscount
> To set it:
> crsctl set css misscount 45
> "crsctl get css" will give you other possible settings that are
> Note that Oracle advise to not set misscount to a value lower than 30
> Hope that helps,
> On Mon, Feb 6, 2012 at 2:36 PM, Purav Chovatia <puravc_at_gmail.com> wrote:
> > Is there any crsctl (crsctl status resource ora.net1.network -f does not
> > show) or srvctl or any other command to display this timeout?
> > Is this timeout modifiable? i.e. can the VIP be made to failover sooner
> > the surviving node?