Start a Conversation

Unsolved

This post is more than 5 years old

2103

March 24th, 2011 13:00

Query Order on Second Centera

As I understand it, query returns clips in order based on the physically stored date of the clip.

If we consider a single Centera (and no "funny" stuff like FPClip_RawOpen), that means a query will return clips in chronological order as they were written by the application.

Now let's add a second replica or replacement Centera to the picture, and populate it with all the clips from the original Centera.  I guess this could be accomplished with Centera replication or the CASCOPY utility.  (Are there any other methods without getting the application involved?)

Would a query on the second Centera still return clips in the same chronological order as they were written by the application to the original Centera?

I understand that replication will give the replicated clips a fresh physically stored date.  If replication copies clips in physically stored date order, the same order will be retained on the new Centera, although with different physically stored dates.  But if replication copies clips in any other order, the clips on the new Centera will be "out of order" with respect to the original Centera.

I don't have any clue how CASCOPY works, so I won't make any guesses about it.

To summarize my question, under what circumstances after using Centera replication or CASCOPY will the query order on a replica or replacement Centera be the same as on the original Centera?

Thanks,
Jim

2 Intern

 • 

417 Posts

March 24th, 2011 13:00

As query is based on the time when the clips arrived on the cluster, not the metadata within the clips (such as creation.date or mofification.date) then the only time the order will be the same would be if the clips were replicated (or written) in exatcly the same order, or a CentraStar restore operation was performed using a backup of the original cluster.

I guess that should pretty much cover your question!

208 Posts

March 28th, 2011 05:00

Hi Jim -

Just to strengthen Graham's point:  In real customer scenarios with millions of clips and years of runtime in a Centera application environment it is nearly guaranteed that the ordering of query results between replicating clusters will not match, even if those clusters are perfectly "synchronized" as far as their contents are concerned.

Best Regards,

Mike Horgan

http://www.interlock-tech.com

No Events found!

Top