Re: EMC storage replication

  • From: Fuad Arshad <fuadar@xxxxxxxxx>
  • To: William Muriithi <william.muriithi@xxxxxxxxxxxxxxxxxxx>, "goran00@xxxxxxxxx" <goran00@xxxxxxxxx>
  • Date: Thu, 2 Dec 2010 08:15:12 -0800 (PST)

That is absolutely true.
Replication is done at a block level and any corruption on the primary on the 
disk level will be passed and not noticed until the array is used. 

but it depends on what technology is being used. i'm assuming SRDF/A will be 
used since its between 2 different datacenters. there are ways to put delays to 
ensure that all blocks are nto replicated immediately.




----- Original Message ----
From: William Muriithi <william.muriithi@xxxxxxxxxxxxxxxxxxx>
To: "goran00@xxxxxxxxx" <goran00@xxxxxxxxx>
Cc: Fuad Arshad <fuadar@xxxxxxxxx>; "centroid.dba@xxxxxxxxx" 
<centroid.dba@xxxxxxxxx>; "JEREMY.SHEEHAN@xxxxxxxxxxxxxxxxx" 
<JEREMY.SHEEHAN@xxxxxxxxxxxxxxxxx>; ORACLE-L <oracle-l@xxxxxxxxxxxxx>
Sent: Thu, December 2, 2010 10:08:14 AM
Subject: Re: EMC storage replication

Goran

This is not from personal experience - Have only worked with DRDB which is the 
same animal in different skin, but I do recall reading it somewhere and sounded 
like something that can be a deal breaker.

EMC has no application knowledge and replicate everything in block level. 
Essentially, every change that happen on the primary array will be pushed to 
the 
secondary array even if its undesirable. So, a corruption/misconfiguration on 
your primary database would not only bring down your primary database, but you 
would shortly get alerts from scripts monitoring the backup database that all 
things are not fine there too. Dataguard  would however not replay corrupted 
logs, the replication would just fail when oracle detect that the logs have 
issues and hence save the day.

So yes, dataguard may be expensive to license and maintain that EMC 
replication, 
but I would not look at one as a complete replacement of the other.  This is 
straight from my memory, which from experience is not that great, so research 
on 
it and update us should I be misinforming the list


William

On 2010-12-02, at 10:16 AM, goran bogdanovic wrote:

> Thanks you all for answers!
> 
> The idea behind is following:
> On primary Data Center we will be having 'primary' database attached to EMC 
>SAN.
> All disk changes should via EMC Replication sent to SAN in second Data Center 
>located in different town (I have been assured by network guys we have enough 
>bandwith for this).
> On second Data Center we should be having 'secondary' (kind of 'standby') 
>server connected to 'replicated' SAN - 'active - passive' configuration.
> 
> In case that 'primary' server or Data Center 'fails', we should be able to 
>activate 'secondary' server and "Le voilà !" ... the system is running again 
>;-)
> 
> sound simple and EMC sales state it works ;-) ... and I always double check 
>statements comming from sales people.
> 
> @Jaffar,
> convincing management is not an option ;-)
> 
> 
> Best Regards,
> Goran
> 
> 
> On Thu, Dec 2, 2010 at 3:55 PM, Fuad Arshad 
><fuadar@xxxxxxxxx<mailto:fuadar@xxxxxxxxx>> wrote:
> Goran,
> We are using both SRDF/A and Server Cloning Processes for replication.
> The process is fairly simple  with the most of the coding/scripting  effort .
> One of the Issues we've encounted is while the Clone process is running There 
>is a very high amount  of I/O Being Done as the Disk Frame level. Since a 
>clone  
>can only be done in the Same Frame level. This can cause Database Bottlenecks 
>albeit for a small period of time .
> We've Move to Data guard  due to these issues.
> May i Ask what  is the requirement .  If Offloading Backups is the need. I've 
>found Physical Dataguard to be better than  EMC Cloning.
> SRDF/A is a different product set from EMC and it allows for between 
>Frame/Datacenters replication.  It is a fairly costly solution though.
> While Oracle will certify an Oracle Database Running on an EMC Frame. It will 
>not certify a 3rd Part product.
> The technology is called Split mirror  and the EMC Product is Time Finder.
> Meta link Note Exploiting EMC Timefinder and Oracle Recovery Manager [ID 
>204719.1]
> 
> 
> 
> ________________________________
> From: goran bogdanovic <goran00@xxxxxxxxx<mailto:goran00@xxxxxxxxx>>
> 
> To: ORACLE-L <oracle-l@xxxxxxxxxxxxx<mailto:oracle-l@xxxxxxxxxxxxx>>
> Sent: Thu, December 2, 2010 6:58:54 AM
> Subject: EMC storage replication
> 
> Hi all,
> 
> at the place where I work, management would like to implement EMC Replication 
>(as we have EMC SAN's) for our databases between two Data Centers instead of 
>DataGuard for certain DB-systems.
> According to EMC sales, the process is certified by Oracle although I didn't 
>find any reference on Metalink (could be also, I didn't search it well).
> If someone (out there ;-) ) is already using it, I would appreciate if you 
>could tell me your experiences and possible pitfalls regarding implementation.
> 
> Best Regards,
> Goran
> 
--
//www.freelists.org/webpage/oracle-l


Other related posts: