Data ONTAP 8 takeover-giveback information

[ThomasBebee]

From the presentations I have seen, it appears that the takeover-giveback process in Data ONTAP 8 will be a significant improvement over the 7.3x process.  I have been trying to find some additional detail on what these improvements are, but have not found anything yet.  If someone is aware of any write-ups on this topic, could you please point me to where it exists, as I would like to learn more of the specifics.

Thank you.

Tom

[amiller_1]

Nothing publicly I'm aware of I'm afraid.....I'd check with your NetApp or partner sales engineer to see if they have info they can pass on informally or under NDA....

[droach]

Don't know of any docs, but here are some results of my testing.  Testing to see how ESX and its clients react to takeover and giveback.

• nSeries 7900 active/active configuration
  
• OnTap v8.0 7-Mode
  
• ESX - vSphere 4 Enterprise Plus
  
• Initiator group configured with ALUA on.
  
• ESX host with two single port Qlogic QLE2460 HBA's
  
• ESX host using Round Robin and Storage Array Type:WMW_SATP_ALUA
  
• ESX host zoned to both Heads - ESX automatically and correctly identified its 2 primary paths (Head-A), and 2 secondary paths (Head-B) (see insert below)
  
• Using IOMETER on Windows guest to generate I/O to SAN-attached LUNS
  
• Using PERFMON on Windows guest to monitor I/O
  

Taking over Head-A - issuing takeover command FROM Head-B

• Perfmon shows a 10 second loss of connectivity (no I/O) to its LUNS on Head-A
  

Issuing giveback -f command FROM Head-B

• Perfmon shows a 30-40 second loss of connectivity (no I/O) to its LUNS on Head-A
  

This is the first time we are using a Filer strictly for FC-attached back-end storage.  I am not sure if this loss of connectivity is considered "normal".  Does anyone have any suggestions to eliminate this connectivity loss during takeover and giveback?  We are seeing similar results for physical Windows servers.

[masaru_ryumae]

Sorry, this is a late reply, but we are also seeing similar issues of losing paths lately. I don't think we saw this before, but we are now seeing it (or we have been seeing this, but I didn't pay fully attention to notice it).

I am told by NetApp support to make sure clients (ESX/i, Linux/Windows) have Host Utilities (5.3 or 6.0 based on support compatibility matrix), and multipath configurations are fully correct. I will be going through my client systems to make sure clients are fully compliant, but have you tried host utilities with proper multipath config for ALUA? Does lun check_config -A show FC clients with partner paths used regularly?

If you have used host utilities, has it helped to handle well during fail-over/giveback?

I will report out what I find out on our next fail-over/giveback opportunity.

Thanks!