we have a really strange netapp snap mirroring issue that came up after me made some changes to the nexus 7000's in our data centers. we have 2 data centers. dc1 has 2 n7k's and dc2 has 1 n7k. we had them connected thru a single 10gb link. a few weeks ago we added a second link and setup the links as a port-channel. vpc on dc1 end and regular po at dc2. we also have 2 netapp controllers in dc1 and 2 in dc2. after we did this the snap mirroring became very slow form 1.2 gb to 90mb. but this just occured when it went over 1 of the data center links using dc1 n7k-2. we tried shutting down and reversing the links but the problem seemed to follow the dc1 n7k2. we then frorced everthing over dc1 n7k-1 and we had the same slowness issue. we then decided to move the connect in dc2 n7k-1 from slot 4 to slot 1 (where the other connection is) and it resolved the issue. both modules are m1 10gb series cards (no differences between the cards). we had already opened a tac case with cisco (626790461 : slow data transer between n7k's using 10gbase zr optics). they send us a new module for slot 4. we installed it and the same problem. we moved the module to slots 3, 7, 8 and the same problem. the only way we can get this to work is to have the port-channel in dc2 have both interfaces on the module in slot 1. we also opened a case w/ netapp and they are saying it's a network issue. we have tried everything we can think of to resolve this issue and nothing seems to work.
it works without the bandwidth commands. when both interfaces in the port-channel are connected to module 1 (they are both 10gb connections). when we move one interface to another module the speed is 1/10 of what it was. there are no diferences between the two modules that i or cisco tac can find. the configuration doesn't change. the source and destination host connections don't change. the only thing that changes is the port-channel between the data centers and the change is only on the data center 2 side.
yes, so far it only happens with netapp and we have jumbo frames enabled. we actually tried removing jumbo frames and it didn't help. we have not tried connecting a server to these ports and trying to replicate the issue.
the problem seems to be that the netapp controllers are seeing something different when determining how fast to send the data, depending on what the path is. but the only difference in the path is the module the port-channel between data centers connects to. any ideas would be appreciated. thanks.
Learn, share, save
Discover and save your favorite ideas. Come back to expert answers, step-by-step guides, recent topics, and more.
