UNSOLVED

Jsang2

updated

15 years ago

J

Jsang2

16 Posts

0

2359

June 23rd, 2011 07:00

Mirrorview/A over iscsi

I have 2 clariions CX4-120 , 1 at live site and the other at disaster recovery site , Mirrorview/A is enabled , There are  11 LUNS  to be replicated to  disaster recovery site through iscsi  over a  2mb/s link  , 6 luns are  1TB and the other 4 are 2 TB , The problem is that it is taking too long to synchronise  , On fracturing some luns  leaving 2 to do initial synchronise  , it moves 1%  after 2 days , What could be causing this slowness

  • Jsang2

    16 Posts

    972

    0

    Posted June 23rd, 2011 07:00

    Forgot to mention that latency between the iscsi ports on the 2 storage nodes is very high about  1000ms and keeps going up

    Trace root has double hops

  • yassine2

    64 Posts

    972

    0

    Posted June 23rd, 2011 07:00

    have you tested the line with ftp for example?

    have you concistency groups?

    have the initial synchronisation been successfull?

  • kelleg

    6 Operator

    4537 Posts

    972

    0

    Posted June 23rd, 2011 11:00

    Have your network provider determine the bandwidth, not what it is suppose to be, but what is measured. 1000ms latency is pretty bad. Also, a 2mb (bit or byte) is slow - that sounds like a T1 connection speed 1.5Mbit /s (150KBytes/s).

    glen

  • Jsang2

    16 Posts

    972

    0

    Posted June 24th, 2011 08:00

    Ftp is blocked on the network

    At the moment I don't have any Consistency group

    1 luns is running the initial synchronisation

    Are there any tools from EMC that can be used to measure bandwidth utilization ?

  • DanJost

    190 Posts

    972

    1

    Posted June 27th, 2011 07:00

    When I implemented MV/A, EMC came out and tested the circuit.  I'm assuming you/they never did this?  I am running MV/A over the iSCSI ports without any problems but my WAN circuit is 100Mbits.  When they tested it they "certified" up to 85Mbits and did the feasibility calculations based on that.  The result was that we could MV/A the entire storage system on a fairly short time schedule.  Your circuit seems awfully slow for the amount of data you are trying to synchronize.  You will get high latency when the circuit is essentially 100% full which is probably happening on the initial sync so unless you have other network issues a high latency might not be the problem only a symptom of not enough bandwidth.

    You'll also run into problems if you are a frame-relay like guaranteed bandwidth setup with best-effort burst.  I don't see many circuits like this anymore, though...but if you do you have the potential for a lot of discards and retransmits.

    Dan

  • Jsang2

    16 Posts

    972

    0

    Posted July 15th, 2011 00:00

    Thank you all for the answers given, The network was identified to be cause of the problem , Network link was upgraded to  3 MB/s and seems faster now however the Lun which was doing the initial synch but the the time didn't change on the Lun  and the image state is showing (T) synchronizing

    Last started is showng the date when the remote mirrorview/A was created and  Last ended is showing  N/A

    The other question is the incremental copy seems to restart any time the link goes down , Is there any setting I missed out ?

  • christopher_ime

    6 Operator

    1962 Posts

    972

    1

    Posted July 23rd, 2011 20:00

    >> The other question is the incremental copy seems to restart any time the link goes down , Is there any setting I missed out ?

    Please refer to the following post which will explain what you are seeing.

    https://community.emc.com/thread/123610