Slow Replication Using Replication Manager

Question

These are 500gb VMFS LUN's, replicating from an NS20 to an NX4. Currently, the Celerras are sitting right next to each other behind routers (NS20 is production, NX4 will be a portable DR site). The replication manager servers are VM's running Replication Manager 5.2.x on ESX servers. Each site has gig routers and switches, but the replication is taking very long. I have successfully replicated and mounted a 1gb VMFS LUN, but am having difficulty with the larger LUN's. I can imagine that this would be slow over a WAN link, but when monitoring the router interfaces, there is hardly any traffic going over the line (consistently 30Kb/s, up to a few MB/s rarely). I don't currently have any data mover throttling setup. Is there something that I am missing? Thanks!

JamesBEMC · Answer

Hi MarkCan you tell me a bit more about how the job is configured in terms of Celerra replicator?Also, have you tried to manually configure a Celerra replicator job and see how that performs as opposed to one under RM's control? I suspect the issue would be the same (that is of course without knowing more).You also say you managed to replication a 1GB VMFS Lun, but are having difficulty with larger LUNs. Did the 1GB LUN replicate with the desired speed (much faster) or as it was much smaller, it was replicated slowly but got there in time?ThanksJames

marksoho · Answer

The replications are setup to use IP replicator. They are using, each, an interface that is either FSN (ns20) or lacp (nx4) for the replication (CGE1 and 3 on both). The latest replication attempt (did not finish) reached 38% and then just stopped.

In the DR site Repman Server, I received this error:

000024 ERROR:An unexpected internal error occurred: bad allocation.

The XML message for above-mentioned error is as follows:

final

2009 02 01 09:30:01

InternalError
bad allocation

Are there any specific items that I need to take care of when the Rep Manager servers are VM's? Thanks again! The replication still is very slow, only bursting up to 25mb/s rarely and briefly.

JamesBEMC · Answer

Hi MarkI dont think VMs are an issue here.Can I suggest as part of a troubleshooting/isolation test, can you run a manual (outside of RM) replicator test of the same LUNs and see if the performance / stability is the same or if it is significantly better?ThanksJames

marksoho · Answer

I have isolated the issue to the routers. I setup a VLAN on layer 3 switches and the replication is running perfect across them. The routers are running JUNOS Enhanced Services, which could be described as a firewall. I will do some more research regarding the ports, etc, that need to be open across the WAN, but does anyone have that information offhand? Thanks!

JamesBEMC · Answer

Cool stuff!You ask, I retrieve http://powerlink.emc.com/km/live1/en_US/Offering_Technical/Technical_Documentation/300-004-184.pdfPage 22 - System requirements for Celerra ReplicatorIP addresses configured for the source and destination Data Movers (ports 8888 and 8887 used by replication for transferring data and internaloperations - contact Customer Service to change this setting).¿ HTTPS connection between the Control Station on the source site and theControl Station on the destination site (port 443 - cannot be changed).Note: ICMP (Internet Control Message Protocol) ensures a destinationCelerra Network Server is accessible from a source Celerra Network Server.TheICMP protocol reports errors and provides control data about IP packetprocessing.Also EMC KB emc1711448887 TCP RCP Yes Used by Celerra Replicator during a target to source resync. This port must be open between the source and target Data Movers. 8888 TCP RCP Yes Used by Celerra Replicator. This port must be open between Data Movers on both sides of a replication session. CheersJames.

Replication Manager

Was this post helpful?