RE: Regarding RAC vip failover and tnsnames entry
- From: "Crisler, Jon" <Jon.Crisler@xxxxxxx>
- To: <riyaj.shamsudeen@xxxxxxxxx>, "Niall Litchfield" <niall.litchfield@xxxxxxxxx>
- Date: Mon, 11 May 2009 15:40:06 -0400
Agreed that he needs to fix up his Oracle client tnsnames, and make sure
it is a recent version with patches (say 10.2.0.3 or 10.2.0.4 assuming
10g version).
If you are trying to segregate work to specific nodes, or only use the
2nd node in the event of 1st node failure, then modifying the 'services'
(perhaps with DBCA) to set preferred and available attributes would be a
better way. If your application can use TAF, then you should get
seamless failover even if the 2nd node is not a preferred node.
________________________________
From: oracle-l-bounce@xxxxxxxxxxxxx
[mailto:oracle-l-bounce@xxxxxxxxxxxxx] On Behalf Of Riyaj Shamsudeen
Sent: Monday, May 11, 2009 2:48 PM
To: Niall Litchfield
Cc: piontekdd@xxxxxxxxx; sjaffarhussain@xxxxxxxxx;
sanjeevorcle@xxxxxxxxx; oracle-l@xxxxxxxxxxxxx
Subject: Re: Regarding RAC vip failover and tnsnames entry
As Niall pointed out, OP is running active/passive nodes. So, he needs
to have two addresses in that tnsnames entry and more importantly,
load_balance switched off.
If connection time load balance is on, then one of the address is
psuedo randomly chosen, which is probably not preferred in OP's case. He
needs to have both addresses in that tnsnames string, load balance
switched off and failover switched on.
Cheers
Riyaj Shamsudeen
Principal DBA,
Ora!nternals - http://www.orainternals.com
Specialists in Performance, Recovery and EBS11i
Blog: http://orainternals.wordpress.com
On Mon, May 11, 2009 at 1:33 PM, Niall Litchfield
<niall.litchfield@xxxxxxxxx> wrote:
And to add a bit. The OP described using RAC effectively to mimic an
active/passive cluster. There are cheaper and better solutions for
that.
On 5/11/09, Bradd Piontek <piontekdd@xxxxxxxxx> wrote:
> Not to sound contradictory (although I am contradicting what Riyaj
wrote),
> but the 2nd node is note 'required' to be in the tnsnames.ora if
Server-Side
> Load balancing is in place and you are using the VIP address. All
listeners
> on all nodes know of the other listeners, and can re-route traffic to
a
> different node. What is required is that the DNS entry spit back to
the
> client from the listener needs to be able to be resolved by the
client.
> As with just about anything, especially with RAC, it is important that
you
> test out all these types of scenarios and see how your application
reacts to
> these types of events. (regardless of what the manual says 'should'
> happen). Just $0.02.
>
>
> Bradd Piontek
> "Next to doing a good job yourself,
> the greatest joy is in having someone
> else do a first-class job under your
> direction."
> -- William Feather
>
>
> On Mon, May 11, 2009 at 12:53 PM, Riyaj Shamsudeen <
> riyaj.shamsudeen@xxxxxxxxx> wrote:
>
>> Hi Sanjeev
>>
>> >> Is it required to have second node entry in tnsnanames.ora for
>> connection to work
>> >> after vip failover.
>> Yes, you need to have second node entry also in your tnsnames.ora
file.
>>
>> You might want to understand what problems VIPs are resolving. Let's
say
>> that a new connection request is made and for some reason IP address
in
>> that
>> tnsnames entry is not available and so, there is no response from
that IP.
>> Connection request will wait for TCP connect timeout parameters
(typically
>> 3-10 minutes controlled by tcp parameters), then try next entry in
the
>> tnsnames list.
>>
>> By failing over VIP to surviving nodes of the cluster, there will be
an
>> immediate response, essentially as, 'no listener'. So, connection
request
>> will try next entry in the tnsnames entry immediately avoiding tcp
connect
>> timeout delays.
>>
>> HTH
>>
>> Cheers
>>
>> Riyaj Shamsudeen
>> Principal DBA,
>> Ora!nternals - http://www.orainternals.com
>> Specialists in Performance, Recovery and EBS11i
>> Blog: http://orainternals.wordpress.com
>>
>>
>>
>> On Mon, May 11, 2009 at 12:08 PM, Syed Jaffar Hussain <
>> sjaffarhussain@xxxxxxxxx> wrote:
>>
>>> Sanjeev,
>>> When a node fails, the VIP associated with failed node is
automatically
>>> fail over to one of other nodes exists in the cluster.
>>>
>>> The existing connections will typically receive ORA-3113 errors and
the
>>> new connections using address list will select the next entry in the
>>> list.
>>>
>>>
>>>
>>> On Mon, May 11, 2009 at 9:39 AM, sanjeev m
<sanjeevorcle@xxxxxxxxx>wrote:
>>>
>>>> All,
>>>>
>>>> We have two node RAC environment with the applications connections
going
>>>> to first node.
>>>> (No reference to the second node in tnsnames.ora). We are not using
load
>>>> balancing or failover.
>>>>
>>>> If case of nic failure on first node CRS on the other node will
failover
>>>> the vip (raising event and bringing up the resource online on
second
>>>> node)
>>>> and re-arp the machine address on the network.
>>>>
>>>> Question:
>>>>
>>>> (a) Existing connections: They will get an errors saying the
instance is
>>>> down.
>>>> (b) New connections: They will be able to connect to vip on second
node
>>>> (with reference to firsthostname-vip entry in tnsnames.ora).
>>>>
>>>> Are both (a) and (b) true. Is it required to have second node entry
in
>>>> tnsnanames.ora for connection to work after vip failover.
>>>>
>>>> Regards
>>>>
>>>
>>>
>>>
>>> --
>>> Best Regards,
>>>
>>> Syed Jaffar Hussain
>>> Oracle Certified Master (10g)
>>> http://www.oracle.com/technology/ocm/shussain.html
>>> Oracle ACE
>>>
>>>
http://apex.oracle.com/pls/otn/f?p=19297:4:1579866181463918::NO:4:P4_ID:
126
>>> OCP 8i,9i & 10g DBA
>>> RAC Certified Expert
>>> Official Oracle RAC SIG Representative for Saudi Arabian region (
>>> http://www.oracleracsig.org/)
>>> I blog at
>>> http://jaffardba.blogspot.com/
>>> --------------------
>>> "Winners don't do different things. They do things differently."
>>>
>>
>>
>
--
Sent from Google Mail for mobile | mobile.google.com
Niall Litchfield
Oracle DBA
http://www.orawin.info
Other related posts:
