So, we had a switch fail today. It failed in such a way that link was maintained on the ports, it just didn't pass any packets.
When this happened, the vif that I had created on our filer did not switch over to the backup link (on another switch). I initially thought it was because I hadn't set it to check to make sure that traffic was actually flowing out the interfaces. However, then I couldn't find anywhere to actually do that.
What's the deal here? Did I totally miss something? How exactly does failover work, anyway?
Thanks, Jordan
Ok, I did some more testing. It really does appear that the NetApp does no verification of an interface's ability to transmit data before using it.
We have Summit48i switches, which allow you to disable a port (but leaves the link up on it). I connected both of the filer's interfaces to the switch. One is "e8" (an SC-GigE link), and one is "e0" (the built-in 10/100 port on the filer).
I had the vif set up with neither NIC favored. It was using the e0 interface by default. I issued these commands: vif favor e8 vif nofavor e8
This switched me over to using the e8 interface. All was still well. I was able to ping after switching over to the e8 interface.
Then, I disabled the port that the e8 interface was connected to (leaving link on on that interface). This seemed to have no effect on the filer (well, other than not being able to use the network!). That is, though no data was flowing over the e8 interface anymore, the filer still thought it was ok to have that interface be the active one. I was (of course) unable to ping anything. Then I did this: vif favor e0 vif nofavor e0
Pow. Back on the network, all is well again.
I was completely unable to find any kind of "polling" interface in the documentation for the filer. The other devices with redundant NICs that we have can all poll a gateway via ping, or check for activity via broadcast packets. That way, even if link is still on, the device can detect whether packets are /actually/ flowing across the interface, and failover as appropriate.
Is this truly the case? Can the filer really not tell if the interface is working? I'm a bit surprised, since this seems like a pretty common failure mode (bad port, but link still up).
It fails over just fine if I physically unplug the NIC from the switch, so it's nothing to do with how I have vifs set up or anything.
Help?
Jordan
-----Original Message----- From: owner-toasters@mathworks.com [mailto:owner-toasters@mathworks.com]On Behalf Of Jordan Share Sent: Tuesday, August 07, 2001 5:39 PM To: 'Toasters@Mathworks.Com' Subject: vif's and failover
So, we had a switch fail today. It failed in such a way that link was maintained on the ports, it just didn't pass any packets.
When this happened, the vif that I had created on our filer did not switch over to the backup link (on another switch). I initially thought it was because I hadn't set it to check to make sure that traffic was actually flowing out the interfaces. However, then I couldn't find anywhere to actually do that.
What's the deal here? Did I totally miss something? How exactly does failover work, anyway?
Thanks, Jordan
On the subject, is there any news on VLAN'ed VIF's (rfe/bug 42409)? Adam Fox was able to shed some light a few months ago when I'd brought it up, but I haven't seen anything else officially stated, and the bugid still isn't publically accessible....
I'm about to change alot of stuff and I'm not sure which feature I want more want more.. Mostly I'm hoping that 42409 will be incorporated soon so that I won't have to decide.
Thanks ..kg..
So, I got a reply back from Richard Mccague [richard.mccague@netapp.com], stating that I cannot vif two unlike interfaces. I'm not sure if he meant trunking them or just using them for failover. Failover is working fine in the event of a full loss of link on either interface, so I suspect he was thinking of trunking (where it totally makes sense that you'd not be able to trunk two unlike interfaces).
We do have a second gigabit NIC in the NetApp; we just don't have a second GBIC to plug it into, so we were using the 10/100 interface. We can easily purchase another GBIC, but I want to make sure that I will get "proper" failover in that setup (i.e., two Gbit interfaces).
Does anyone have any experience with this? Can someone confirm or deny the failover in the event of loss of dataflow, but not loss of link?
Thanks, Jordan
-----Original Message----- From: owner-toasters@mathworks.com [mailto:owner-toasters@mathworks.com]On Behalf Of Jordan Share Sent: Wednesday, August 08, 2001 12:51 PM To: 'Toasters@Mathworks.Com' Subject: RE: vif's and failover
Ok, I did some more testing. It really does appear that the NetApp does no verification of an interface's ability to transmit data before using it.
We have Summit48i switches, which allow you to disable a port (but leaves the link up on it). I connected both of the filer's interfaces to the switch. One is "e8" (an SC-GigE link), and one is "e0" (the built-in 10/100 port on the filer).
I had the vif set up with neither NIC favored. It was using the e0 interface by default. I issued these commands: vif favor e8 vif nofavor e8
This switched me over to using the e8 interface. All was still well. I was able to ping after switching over to the e8 interface.
Then, I disabled the port that the e8 interface was connected to (leaving link on on that interface). This seemed to have no effect on the filer (well, other than not being able to use the network!). That is, though no data was flowing over the e8 interface anymore, the filer still thought it was ok to have that interface be the active one. I was (of course) unable to ping anything. Then I did this: vif favor e0 vif nofavor e0
Pow. Back on the network, all is well again.
I was completely unable to find any kind of "polling" interface in the documentation for the filer. The other devices with redundant NICs that we have can all poll a gateway via ping, or check for activity via broadcast packets. That way, even if link is still on, the device can detect whether packets are /actually/ flowing across the interface, and failover as appropriate.
Is this truly the case? Can the filer really not tell if the interface is working? I'm a bit surprised, since this seems like a pretty common failure mode (bad port, but link still up).
It fails over just fine if I physically unplug the NIC from the switch, so it's nothing to do with how I have vifs set up or anything.
Help?
Jordan
-----Original Message----- From: owner-toasters@mathworks.com [mailto:owner-toasters@mathworks.com]On Behalf Of Jordan Share Sent: Tuesday, August 07, 2001 5:39 PM To: 'Toasters@Mathworks.Com' Subject: vif's and failover
So, we had a switch fail today. It failed in such a way that link was maintained on the ports, it just didn't pass any packets.
When this happened, the vif that I had created on our filer did not switch over to the backup link (on another switch). I initially thought it was because I hadn't set it to check to make sure that traffic was actually flowing out the interfaces. However, then I couldn't find anywhere to actually do that.
What's the deal here? Did I totally miss something? How exactly does failover work, anyway?
Thanks, Jordan
-
Jordan Share -
kevin graham