Slow Initial Mirror Replication (asynchronous)

Question

Hi all,

We have recently implemented a VNX5300 Block (production site) (software ver. 05.32.000.5.201) and am now Mirroring/A 5 LUNs 2.7TB (4 x 500GB, 1 x 700GB) to a CX4-120 (DR Site) (software ver. 04.30.000.5.525) the issue it that the initial sync is taking an age, it’s been running for a week and they are only around 15% complete.

The network is as follows, 75mbps link between the sites, the sites are about 5 miles apart. The ISP’s equipment won’t support Jumbo
frames, MTU on SP’s set to 1500. Ping time between SP’s is 1-2ms. All SP’s can ping each other.

I have Free Reserved LUNs 10 x 60GB (20%).

The VNX has replaced a CX3-20 that was doing the same job and mirrors were in sync ok (not sure back in the day how
long that took to sync).

I know the first thing you should look at is network equipment end to end but can’t find any issues there, all equipment (ours or ISP’s) are running at least 100mps.

Anything else I should be looking at or should I expect it to take weeks for my initial sync?

Should I have only created one mirror at a time wait for 100% sync and create the next, I thought it would be able to
handle all 5 at once?

Thanks

kelleg · Answer

Initial sync can take a long time - if you're link is 7.5MB/sec at theoretical speed that would be about 100 hours (2.7TB / 7.5MB/s). There is a sync rate (medium is the default) for each mirror that you can check to see if it could be increased and there is the issue of the speed of the mirror LUNs on the DR array.

In the attached Mirrorview KNowledgebook there is a section that talks about the speed of the link and gives some pointers as to the expected speed for different links. This document and more are available at the support.emc.com portal under the Support By Product section.

glen

Rainer_EMC · Answer

Besides link speed doesn't give you the whole picture - latency also plays a role in what thruput you can get

You might want to look at a WAN optimization device

Or a replication technology like RecoverPoint that can deal better with WAN links

r0uhelp · Answer

Thanks for the replies, I'm figured out the issue and it is now sync'ing at the expected rate.

Our ISP was able to tell me there was a constant stream of traffic over the WAN but it was only using 5mbps of bandwidth so I started by checking end to end nodes to find where the throttle/bottleneck was.

I noticed even though on the production side the iSCSI ports on the VNX were set to 1Gbit I could see by the link lights on the switch where they are terminated to this was not the case, I rebooted both SP's (one at a time), link light on the switch indicated 1Gbit connection and the WAN traffic shot up to full bandwidth 75Mbps.

Bit of a strange one but at least it's now working ;-)

Rainer_EMC · Answer

Thanks for the feedback It would have been interesting what the interfaces negotiated

VNX

Slow Initial Mirror Replication (asynchronous)

Was this post helpful?