netapp FCP failover

[RAKESHMOLE2]

I have fas8020 netapp in HA 7 mode.

I have FCP connection shared through two san switches , i was testing the HA failover , i setup mulitpaths on the luns.( it sees 4 paths two from each controller)

However i notice something strange that when i fail over two paths fail which i assume that is the path assosciated with the contoller that failed over.

That shouldnt be the case though isnt it , as the take over controller should take over the fcp paths from both the failed over controller.

I assume it may be the zoning.

[aborzenkov]

What you observe is correct. Partner takes over serving data on behalf of failed controller, but paths to failed controller are no more available.

[RAKESHMOLE2]

hi aborzenkov ,

So you are saying this is expected behavior and this is what i believed all this while too , the reason why i am doubting this is because i have a fas3210 through which i had presented a fc p lun to a server where i set no mulitpaths , the fas3210 is in ha 7 mode and when i fail over those controllers the fcp traffic is seamlessly moved over.On the fas8020 i had to set up multipath I/O for the lun traffic to be not to be disrupted , but on fas3210 i had to do no such thing it automtaically failed over.

the only difference between them is on fas8020 the fcp traffic ports are from onboard fcp port and pcie card and one the fas3210 they are both on onboard ports.

is there a document i can refer to about fcp failover happens in details?

[DOMINIC_WYSS]

you always need multipathing on the host. in Windows you need to install MPIO, then it sees all pathes and does seamless failover.

however, two of those four pathes are on the wrong head. so it is recommended to install Netapp DSM, which is a plugin for Windows MPIO.

then it will see which pathes are on the correct head (optimized) and which are on the 'wrong' one (non-optimized). in case of a failover, it will

switch to the non-optimized ports while the optimited ones are offline.

you need to configure ALUA on the iGroup for this to work.

a VMware ESXi host can do MPIO with ALUA by itself, no need to install anything.

[aborzenkov]

Windows server starting with 2008 has native support for ALUA and does not need NetApp DSM as well.

[DOMINIC_WYSS]

it does not NEED it, but imho it's better with Netapp DSM.

two advantages:

- Microsoft  does per default RoundRobin, so it will send traffic over the non-prefered path (which will generate ASUPs with "partner path misconfigured"). I've never seen it doing really ALUA...

- you have a GUI where you can quickly see which paths are online, which are prefered/non-prefered pathes, etc.

[RAKESHMOLE2]

hi ,

I understand the multipath recommendations that you guys are referring to. The only baffling part for me and has me confused is that on my fas3210 on the fcp connection to a citrix xenhost where i had no multipath set up , when i did a failover the failover was seamless between the two HA 7 mode controllers, but on my FAS 8020 ( same HA 7 mode) when i did a failover initially without mulitpath enabled the Citrix xen host lost access to the luns and when i setup multipath and i did a failover it lost access to two paths ( which i assume is the paths to the failed controller) which should be logical,

I am baffled by the behaviour of my FAS3210 of how could it keep the connectivity to the luns with no mulitpath/no alua enabled ( both on iSCSI conections and FCP connections)