Reconfigure CX4-480 so Windows/Linux server can see all volumes as local drives

you can't do that, Celerra uses proprietary file system uxfs

http://www.scribd.com/doc/140152646/01-07-09-Storage-Virtualization

Just to clarify, I don't want to use the Celerra Control station as the server that needs to see the volumes. I am willing to eliminate the control station out of the equation to have another server(s) see the volumes.

We have another EMC system (CX3-10c) setup as SAN where there is no control station, but rather 2 active/passive servers that are connected to the storage through FC switches.

I also remember the EMC engineer mentioning that this system can be reconfigured to SAN.

if your goal is to connect a Linux/Window server to CX4-480 and give it access to LUN currently presented to Celerra datamover, then it cannot be done. I mean you can, but you will see garbage and will most likely corrupt uxfs file systems. If you just want to connect another server to CX4, create new LUNs and present them to this server ..i see no issues with that.

I want to connect another server to CX4, create or rather recreate the existing LUNs and present them to this server.

the data currently present on the LUNs is not needed, so I pretty much have a fresh shot at the system without worrying about losing anything.

How can I accomplish this?

Thank you.

do you know what model of HBA ? These are supported

if you are going to connect directly, you will need to play with HBA setting to get them to login to CX4 (might have to change topology to "point to point" or something of that nature)

I have the 41ES and the 42ES. So Do I connect the SPA to the new Windows server and set the topology to point-topoint. What else do I need to do? Is there anything on the Clarrion that I have to do, disconnect the control station?

Please advise.

Thank you for looking into this.

i see, so i assume that you have FC HBAs in your servers ? Do you have fiber channel switch(s) or did you plan on connecting these servers directly ?

i assume that you have FC HBAs in your servers -> servers? Are you asking about SPA & SPB or the server I want to attach to the storage? SPA and SPB both have extra ports on their HBAs. I do  have an HBA on the server I want to attach, however I'm not certain if the card is supported. I have few ATTO Celerity cards, I hope they are supported.

Do you have fiber channel switch(s) or did you plan on connecting these servers directly-> I have 1 switch, but I don't mind connecting them directly.

Thank you.

I will need some time to get this setup, hopefully towards the end of the day. Once done, I will get back to you.

Thanks again for helping with this.

Good news,

I configured a Windows 2008 server with a single HBA (2 ports; one for SPA and 2nd for SPB). Like you predicted I had to change the Topology settings: from default point-to-point to Arbitrated Loop for the cards to login to Clariion.

I configured a Storage Group, added 10 LUNs and the newly added Host. The host sees 10 separate LUNs so I spanned them all together on the OS level to create one large Drive ~18TB.

This is just a test system before I configure the real thing so I have few follow up questions:

1) When I manually registered the two fiber ports I had to chose "Initiator Type". I chose Recovery Point Appliance!

the available choices are: Crariion Open, HP Auto Trespass, HP No Auto Trespass, SGI, Fugitsu Siemens, Compaq /Tru64, and Recover Point Appliance. I also chose Legacy failovermode 0 as Failover Mode. What would be the best practice for the Initiator Type and Failover mode?

2) It appears some LUNs are owned by SPA and some by SPB, should I chose half LUNs from each SP for the host to ensure redundancy or does it even matter?

3) Is there a way to present one large LUN to the host instead of many smaller ones to avoid spanning them on the OS level?

4) Do I need PowerPath or some other software to ensure failover works as needed?

Dynamax, thank you very much for guiding me this far.

Make sure to configure multipathing software on the host, since this is a test system i don't think you will be investing into PowerPath, so at least configure the native MPIO. You can find instructions in this doc

Host Connectivity Guide for Windows

1) Select Clariion Open, failover mode 4

2) in terms of performance yes, not really for redundancy ..if one SP fails all LUNs will trespass to its partner SP

3) sure, take a look at MetaLUN functionality ..if you search support.emc.com you will find whitepapers that describe how MetaLUN work and best practices when creating them.

4) answered above

1) i am not sure why HAVT complains, maybe because it only sees one path from each HBA, if you performed your failover test and everything is fine (besides this is test environment), i would not worry about it too much.

2) for general file share server (windows), i would use concatenated MetaLUN. I hate doing host side striping unless i have a very specific requirement where there will be a lot concurrent I/O ..host striping would help there because you could present two LUNs ..one from SPA and one from SPB so both SPs would be handling the workload. For regular "office" file servers, concatenated meta is good enough. Easy to expand, don't have the same strict expansion requirements as striped meta. Make sure partition is formatted as GPT.

Thank you this has been very helpful.

Few more questions:

Even after configuring Multipath I/O and doing a successful failover from SPA to SPB and back to SPA, the Connectivity status for the Windows host shows as "HAVT Issues" ->

There is insufficient information available to determine whether this server (10.X.X.X) is running an approved failover SW package. Please verify that some failover SW, such as EMC PowerPath, is installed and running on the server.

1) Is Multipath not an approved software? I read the doc you send me and it says that my flare(Version: 04.30.000.5.517) should be supported.

I was able to access files on the volume during the failover and right after, so failover seems to work great. No data corruption reported in the event logs as well.

2) I'm in the process of expanding one of the volumes into a very large 17TB metaLUN. It's gonna take forever but I'll wait to see how that works. What method of presenting the LUNs do you recommend for production systems for general file share servers? Span smaller LUNs on the OS level or metaLUNs?

I can't thank you enough for your help with this. I needed guidance with this as the volume of documentation on the subject is enormous. Thank you.

it's my pleasure, enjoy the holidays as well.