vfiler dr failover testing results in duplicate IPs

fletch2007 · ‎2010-09-24

Hi,

I have reproduced this several times in testing vfiler failover with ONTAP 7.3.3

1) (on source cluster) vfiler stop testvfiler (vfiler status reports stopped)

2) (on DR cluster) vfiler dr activate testvfiler@sourcecluster (reports vfiler is activated) - but shortly thereafter "Duplicate IP address 10.6.64.100!! sent from Ethernet address:02:a0:98:xx:xx:xx"

And I did check the KB:

https://now.netapp.com/Knowledgebase/solutionarea.asp?id=kb28734 (it basically says to make sure the source vfiler is reporting stopped before activating the DR side)

What I found was the source side while reporting the vfiler stopped, hangs onto the IP address as an alias

issuing an

ifconfig VIP -alias <duplicate IP>

resolves it - but this feels like a bug and I did not have this issue in versions prior to 7.3.3

Anyone else seen this?

going to open a case too.

thanks

http://vmadmin.info

scottgelb · ‎2010-09-24

I also just noticed this with 7.3.4 and the interface was not an alias. I had to ifconfig down the interface (or set to 0.0.0.0) to remove the duplicate. This has to be a bug since the vfiler is stopped but the physical controller keeps the interface active that was bound to the vfiler.

fletch2007 · ‎2010-09-25

If I had the time I'd run our previous version in the Simulator to verify vfiler stop & vfiler dr activate was indeed working properly in 7.2.4 and not creating duplicate IPs

I'm waiting on Netapp support case # 2001721904 right now - will summarize how it turns out

thanks

http://vmadmin.info

scottgelb · ‎2010-09-25

Very good..then it can get filed and fixed in a P release hopefully. I can confirm it worked fine on 7.3.2, so it broke somewhere at 7.3.3.

I also ran into another issue but with vfiler migrate with 7.3.4. When vfiler migrate moves the vfiler from node A to node B, it correctly stops and destroys the vfiler on node A (after activation on node B), however it also disables snapmirror on node A. If other mirrors are running that will cause issues (easy to fix, but another BURT that needs to be filed).

fletch2007 · ‎2010-10-05

This is the case update I just received - they marked it as closed! - I guess its up to us to demonstrate the behavior changed from 7.3.2 -> 7.3.3?:

"Engineering stated this is working by design and will not be fixed. The vfiler DR feature assumes that dr backup is activated only when source vfiler undergoes disaster. In case of disaster, no resources of vfiler, like ip address,
are accessible. Hence, dr activate does not care about state of vfiler unit at source. It assumes
that the source vfiler unit is not present and continues.

The specified error message will only appear if both the ip's are in the same
subnet. This happens because both the network cards publish, their mac address
to ip address association. As they are in same subnet, there is an ip conflict.
This is a network configuration problem and the workaround is to unbind the ip addess at source vfiler.

I will go ahead and archive this case per the BURT has been addressed."

scottgelb · ‎2010-10-05

This is definitely a bug and not just for vFiler DR, but for stopping a vfiler. When you stop a vfiler the IP stays broadcasting. The behavior did change.

For the other 7.3.3/73.4 issue there is a BURT now for the vfiler migrate bug where the source controller snapmirror protocol is turned off.

mg · ‎2012-02-28

Hello,

This is a known "bug": https://now.netapp.com/NOW/cgi-bin/bol?Type=Detail&Display=451537

It should be fixed in 7.3.7 and is not present in 8.0.x branch due to Data Motion not in it. Bug seems also fix in 8.1.x.

Kind regards,

Michel