Creating dynamic pairs with establish VS. Creating dynamic pairs with invalidate

Invalidating a device does not initiate replication, it only marks the other device as not containing data that is protected and tell the Symmetrix that the tracks on that device can be overwritten by SRDF from another device. Establish initiates replication from a device to another device that is invalidated. Yes, -nowd just allows you to skip the write disabled check

Understand.

However, why do I have to invalidate r2 first, before overwrite it by "#symrdf -sid 55 -file devicefile establish -rdfg 1" command?

Can I just only use first method,  run the single command "#symrdf createpair -file devicefile -sid 55 -rdfg 1 -type R1 -establish" to copy data from r1 to r2. It would achieve the same result, right?

Thanks!

understood your first part.

>but then control when the actual data copy starts.

Could you please explore "the main differences" more? What do you mean by "control"? Why do I need to invalidate r2 first while creating pairs? Without invalidating, I can subsequently run copy data (establish), right?

Under what circumstances, I need to separate steps of "invalidating", and initial full copy? in another words, why do I need to invalidate r2 tracks first, can't I just run 'establish' at once which include both steps?

Including the establish in the createpair command will invalidate the R2 AND start the data copy across the link.  Including the invalidate will only make the R2 data invalid and no data copy will occur.  You will need to run a symrdf establish command later to synch the data across the link.

The main reason to use the invalidate vs the establish is if you want to create the pairs initially, but then control when the actual data copy starts.

emcmagic,

The 'establish' is a compound operation; one of the steps in that operation is to invalidate R2 tracks.

The 'invalidate' is a single-step operation; all it does is invalidate the R2 tracks.

So regardless of which option you use when running the createpair, you are invalidating the R2 tracks. The -establish option invalidates them and then starts performing the initial full copy. The -invalidate option invalidates them, and then it's up to you to perform the initial full copy at a later time.

You cannot run an initial full copy/establish without first invalidating the R2 tracks. Invalid tracks are the way the Symm knows which tracks need to be overwritten, and which tracks are identical. In the case of an initial full establish, you're copying everything from R1 to R2... so we invalidate all R2 tracks first, which tells the Symm it needs to copy everything.

Hope that helps,

- Sean

There probably aren't many circumstances today where you'd want to delay the establish.

But one scenario could be if you had limited bandwidth between sites, and you wanted to execute the createpair during the day... but also wanted to hold off on the establish until a later time, when bandwidth utilization is lower.

I wouldn't say that's the only reason; you might also do it to defer a potential disk/performance impact, or because establish activity is mandated to occur within a maintenance window by your organization. There are other ways to address the bandwidth/disk impact though (e.g. QoS).

If the R2 is a new device, we still need to invalidate the tracks prior to copying data. If both the source & target (R1 & R2) devices are new, the VMAX is smart enough not to actually copy all of the empty tracks over the WAN though. So the initial full copy finishes very quickly between two new TDEVs.

So, in this sense, can I then say the only purpose of the option "invalidate" is to spread the bandwidth? By the way, if r2 is a new assigned lun, which means there are no data on it, then what "invalidate" would do? I am sorry, but obviously, I am slow...