Ordinarily this shouldn't happen. However, can you confirm the followng for me please:
1. Version of RP.
2. Gen type of RPAs.
3. Most importantly, WAN connectivity between the RPA's in each local cluster.
Point 3 is a key point in that the LAN, WAN, iSCSI and Fibre paths between the RPAs act as communication paths as well WAN and Fibre being used as replication pathways. So, irrespective of the replication pathway or replication requirement (local or remote) the RPAs in each cluster must always be connected using these pathways. This ensures that if there is a LAN failure on an RPA, as per your scenario, communication can be maintained between the RPAs via one of the other pathways and site control (RPA controlling the cluster) is not compromised.
SiteA >>>>>>>>>>>>>>>>> SiteB
RPA1 >>>>>>>>>>>>>>>>>> RPA1
RPA2 >>>>>>>>>>>>>>>>>> RPA2
Above is the scenario, RPA1 @ site A is connected with a dedicated link to RPA1 @ site B
Same is with RPA 2 with another dedicated link, all ip’s can ping each other
Now, if LAN cable from RPA1 is removed, CG’s running on RPA1 get into paused status.
Forget about the cross-site communication/inter-cluster WAN relationship. I was referring to the loca or intra cluster WAN connectivity.
RPA1 <-> RPA2 - requires LAN, WAN and Fibre (iSCSI if using vRPA's with VNX)
RPA1 <-> RPA2 requires LAN, WAN and Fibre (iSCS if using vRPA's with VNX)
The replication pathway is a given simply for replication purposes. What I am referring to is the communication pathway.
There is a zone on each switch which has all the RP ports (both RPs at a site) and the storage ports.
How do I make sure that.
All ips can ping each other without any issues.
Also, if we remove WAN cable from a RPA, all its CGs fail over to other local RPA.
Sent from Samsung Mobile
The UI should tell you if you haven't got WAN connectivity between your RPAs in your local clusters. You can also perform a full mesh connectiivty check from boxmgmt to verify connectivity.
If you do have the correct connectivity removnig the LAN cable from the RPA should not induce a pause state on the CGs owned by that RPA.
On this basis I would recommend that you open an SR.