It will depend on the application requesting the update.
As you say, suppose SnapManager for Exchange initiates one update, that update starts, and SME subsequently requests a new update before the first one completes. The controller will fail the request, saying the mirror is not in an idle state. Now it's up to SME whether it wants to wait and retry or simply give up.
I don't know the SME behavior off the top of my head. Protection Manager will wait up to an hour before giving up. Other applications may behave differently.
-- Pete Smoot, NetApp
-----Original Message----- From: Timothy Hollingworth [mailto:thollingsworth@EPLUS.com] Sent: Wednesday, March 23, 2011 9:49 AM To: 'toasters@mathworks.com' Subject: What if SnapMirrror Cannot Keep Up
So if a source/destination pair is doing a snapmirror update and the current update is not complete before a new update is requested, what is the behavior? Does it queue the second snapmirror update and just update it when the first completes?
I understand the behavior when updates are done by the scheduler on the array itself as per snapmirror.conf. In this case it queues the updates but only to a maximum of three.
"If an update is in progress when another is scheduled to occur, SnapMirror starts another transfer as soon as the transfer is complete. However, if three updates pass while the current transfer is in progress, SnapMirror does only one more update; it does not go back and run updates that have been made obsolete by those scheduled later."
But I'm not talking updates initiated by the scheduler on the array itself.
I'm really asking about something like SME performing a backup and requesting an update. The snapmirror update here is functionally equivalent to issuing a snapmirror update command manually. Suppose the mirror does not complete before the next backup and subsequent update request is made (or snapmirror update command issued manually). What if the rate of data change exceeds the ability of network to complete the mirror or if the WAN link to the DR site is just too darn slow? Do the updates queue on the destination and snapshots build up on the source indefinitely? Or do manual or snapmanager initiated updates follow the "queue to three and then give up" rule above?
Anyone know?
Timothy L. Hollingworth Sr. Network Engineer ePlus Technology, inc. 511 Davis Drive Suite 350 Morrisville, NC 27560 678.462.6698 (Direct/Cell) thollingworth@eplus.com
www.eplus.com
The information contained in this electronic transmission and any attachments hereto may be considered proprietary and confidential. Unauthorized distribution of this material to anyone other than the addressed is prohibited. Any disclosure, distribution or use of this transmission for any reason other than their intended purpose is prohibited.