We are having network issues affecting our syncIQ replication between two Isilon clusters (v. 6.5.5). Currently we have one smartconnect zone in each cluster and we are using the SIP of both clusters to replicate. All nodes have LACP configured on their 10gige NICs. We also use the same smartconnet zone for our NFS/CIFS client access.
Is there a way to segregate the replication traffic from all other? Our main purpose is to bypass the network issue we are experiencing by sending our replication traffic through a dedicated network.
Either approach will work, personally I would prefer using VLAN tagging to having to tie up extra switch ports, however be aware that each node can only have 1 active default gateway at any given point in time. So if your production VLAN, let’s say 10, if that subnet’s gateway has a priority of 1, and your replication network’s gateway has a priority of 2, then all traffic will always go out the production gateway. This creates an asymmetric route, where traffic could come in one way but leave another. The way to fix this is that on your replication pool, create a static route to get to your DR cluster. Use syntax like:
isi networks modify pool subnet5:pool5 --add-static-route=192.168.2.0/24-192.168.9.1
that reads as subnet/netmask(of target)-next hop address (aka gateway of this pool).
Thanks Dynamox. Just to make sure I understand, is this what you did:
1. IP all the 1G interfaces on both clusters
2. Add them into a new pool
3. Configure new Smartconnect zone for the new pool
And use the the new pool under the "source node restriction" to run the replication and use the new smartconnect zone name for the destination cluster. is that right?
And use the the new pool under the "source node restriction" to run the replication and use the new smartconnect zone name for the destination cluster. is that right?
We are trying to use a dedicated subnet (non-routable) for synciq and it's failing with with "Source node could not connect to target cluster" error. syncIQ works fine when we use routable subnet. We are using 10gig ports for both the networks with vlan tagging enabled. I created two different smartconnect zones for each subnet.
crklosterman
450 Posts
2949
1
Posted January 17th, 2014 14:00
Dtek,
Either approach will work, personally I would prefer using VLAN tagging to having to tie up extra switch ports, however be aware that each node can only have 1 active default gateway at any given point in time. So if your production VLAN, let’s say 10, if that subnet’s gateway has a priority of 1, and your replication network’s gateway has a priority of 2, then all traffic will always go out the production gateway. This creates an asymmetric route, where traffic could come in one way but leave another. The way to fix this is that on your replication pool, create a static route to get to your DR cluster. Use syntax like:
isi networks modify pool subnet5:pool5 --add-static-route=192.168.2.0/24-192.168.9.1
that reads as subnet/netmask(of target)-next hop address (aka gateway of this pool).
Chris Klosterman, ICSP, ICIE, CCNA, VCP
Twitter: @croaking
Email: chris.klosterman@emc.com
Senior Solution Architect
Offer and Enablement Team
EMC²| Isilon Storage Division