RE: re: connection timeout and nic outage

Good to know.  Thanks for the info and the pointer, Yann.
 
Kind regards,
 
Herta

-----Original Message-----
From: Yann Neuhaus [mailto:[email protected]] 
Sent: Monday, July 02, 2007 22:22
To: Van den Eynde, Herta (TCM)
Cc: [email protected]; [email protected]
Subject: Re: [suse-oracle] re: connection timeout and nic outage


Hi,

Several answers/parameters :-)

As first combine TAF and FAN for JDBC clients is not advised :
http://www.oracle.com/technology/deploy/availability/pdf/MAA_WP_10gR2_Cl
ientFailoverBestPractices.pdf

Normally if the VIP has switched over the select should continue. How
are the "delay" and "retries" configured ?

since 10.2.0.3 there is also a new "connect-time" parameter which goes
faster from one adress to the another in the adress_list, while
connecting :
in sqlnet.ora : 
SQLNET.OUTBOUND_CONNECT_TIMEOUT = 3

You can also in some cases use the "enable=broken" to improve TAF :
      (FAILOVER_MODE =
        (TYPE = SELECT)
        (METHOD = BASIC)
        (RETRIES = 3600)
        (DELAY = 1)
        (ENABLE = BROKEN)
      )

TAF and FAN strongly depends on network configurations.

Best Regards

Yann

---------



[email protected] wrote:
> Probably a bit late, but you can combine TAF with FAN (Fast
Application
> Notification).
>
http://download.oracle.com/docs/cd/B19306_01/rac.102/b14197/hafeats.htm#
> BABECAFD
>
> Kind regards,
>
> Herta
>
> -----Original Message-----
> From: Peter Santos [mailto:[email protected]] 
> Sent: Thursday, May 31, 2007 19:43
> To: [email protected]
> Subject: [suse-oracle] re: connection timeout and nic outage
>
>


Guys,
      In 10g RAC/Linux how should the query timeout be dealt with when
there is a nic failure. I
    understand the concept of NIC bonding, but that is not an option
that we are considering.

    Anyway, my question is how you've dealt with connection
timeouts. In my example, I have TAF
    connection setup with the "BASIC" option. Here is the simple
test case.

        1. sqlplus (using descriptor with TAF configured) into
some instance.
        2. run a select sys_context('USERENV','INSTANCE_NAME')
from dual;
        
             >> go to node from step 2 and do "ifdown eth0"
        3. re-execute query from step 2 .. this should hang ...

    I see the VIP failover almost immediately, but my connection
will hang on that SELECT for about 15 minutes.
    After that it WILL failover to another instance and I can
re-execute the query.

    What I've been trying to get out of Oracle Support is an answer
on shortening the "hang" time. I can't seem
    to make any progress on this ... Has anyone tested this sort of
thing?

-peter

        

    



-- 
Yann Neuhaus
Senior Consultant 
MySQL 5.0 DBA certified
Infrastructure Managed Services

Trivadis AG 
Elisabethenanlage 9 
CH-4051 Basel 

Fax.     : +41-61-279 97 56 
Phone. : +41-61-279 97 55 
Mobile  : +41-79-457 97 56 

[email protected] 
www.trivadis.com