I need to migrate SAN on an AIX 5.3 server from CISCO switch to Brocade

I am trying to determine the steps needed and will appreciate your response

1) verify firmware is compatible between server and Brocade storage switch
2) shutdown Oracle DB
3) varyoffvg
4) switch cables from CISCO to BROCADE
5) ensure SAN connectivity and AIX can see SAN
6) varyonvg
7) start Oracle DB
8) validate DB

If you have 2 or more fiber ports providing multipathing on the server side you might be able to move one leg at a time to avoid or minimize having your DB down.

You don't say what the back end storage is or what multipathing you're using (if any).

1 members found this post helpful.

Hi there,

How do you plan on testing the connectivity after the cabling change?

I'm not sure whether AIX will recognize the disks as being the same devices (hdisks) after the cabling change. My approach would be to remove the disks from the OS first (before moving the cables), with "rmdev". After moving the cables, run "cfgmgr" to detect the disks again.

If you can elaborate on the back end storage, number of paths & multipath drivers as MensaWater requested, we can try and help further.

Good luck!

1 members found this post helpful.

Thank you for your responses.

The disk is EVA 8100. There are 2 FC cards each with 2 ports. I am not able to give any other information at this time until I am able to work with the storage admin and the AIX admin. I will update further soon.

There are 4 FC cards as stated earlier but they are all single path.

2 are connected to the older Cisco which is being decommissioned and they are the SAN disk connections
2 are connected to the brocade fabric and were used for connection to a tape library which is not in use any more but are still zoned

That being said what would it take to use the brocade FC cards for the disk connectivity?


1) shutdown Oracle DB
2) varyoffvg
3) Zone to EVA 8100
4) ????????????????
5) varyonvg
6) start Oracle DB
7) validate DB

Hi,

Without knowing whether a multipath tool is in use, my take would be:

1) verify firmware is compatible between server and Brocade storage switch
2) verify firmware is compatible between EVA and Brocade storage switch
3) shutdown Oracle DB
4) varyoffvg
5) exportvg
6) Remove the hdisks (rmdev -dl hdiskX)
7) switch cables from CISCO to BROCADE
8) Run "cfgmgr"
9) Check that the disks are available (lspv), and check paths if applicable (lspath)
10) importvg
11) varyonvg
12) start Oracle DB
13) validate DB

I hope this helps.

1 members found this post helpful.

Based on this output from the command -- lspath -H -l hdisk10

Does this mean multipath and if so how does that factor into the steps you outlined?


status name parent

Enabled hdisk10 fscsi2
Enabled hdisk10 fscsi2
Enabled hdisk10 fscsi2
Enabled hdisk10 fscsi3
Enabled hdisk10 fscsi3
Enabled hdisk10 fscsi3
Enabled hdisk10 fscsi3
Enabled hdisk10 fscsi3

Hi,

The steps I proposed will work with or without multipathing.

With multipathing (which you seem to have based on the lspath output), you can most likely avoid the downtime by:

- removing the paths through one adapter (for example fscsi2)
- moving that cable to the new switch
- running cfgmgr to add the new path
- check that you have 2 paths again, with lspath
- repeat for fscsi3

As an aside, you might want to reconsider the paths to the disks. The current paths seem unbalanced (3 through one adapter and 5 through the other)?

1 members found this post helpful.

Thank you for your assistance on this cliffordw and MensaWater. I believe our customer is more comfortable with scheduling down time for this activity. These are the steps needed with our multipath environment? One last question, when running cfgmgr will the multipath be done automatically or is there a step (mkpath) needed for that?

1) verify firmware is compatible between server and Brocade storage switch
2) verify firmware is compatible between EVA and Brocade storage switch
3) shutdown Oracle DB
4) varyoffvg
5) exportvg
6) Remove the hdisks (rmdev -dl hdiskX)
7) switch cables from CISCO to BROCADE
8) Run "cfgmgr"
9) Check that the disks are available (lspv), and check paths if applicable (lspath)
10) importvg
11) varyonvg
12) start Oracle DB
13) validate DB

Hi,

Sorry for the late reply. "cfgmgr" will configure all the paths it finds, no "mkpath" required.

1 members found this post helpful.

I really appreciate your assistance. I have enough now to create a POA for migrating AIX 5.3 P5 Plus server from Cisco to Brocade SAN switch

FYI - while rmdev -ld hdiskX will not hurt, it is probably unnecessary. Even the exportvg maybe unnecessary.

Basically, all that is changing is the path variables (in the ODM) and AIX follows a value in he VGDA (the so-called pvid)).

While a very different technology we used to have an exercise for AIX admin class where a local (SCSI) disk would be powered down (rmdev -l hdiskX, no -d) the disk moved to a new SCSI location (new cabinet or just a new slot) and then mkdev -l hdiskX (rather than cfgmgr) and AIX would find the disk in the new location.

In this case - mkdev -l hdiskX might not find a specific disk in new fabric - but cfgmgr - which scans all available ports and paths, should. What I have seen several times is an existing LUN being scheduled to a new zone and being rediscovered as a the same hdisk.

In short, as long as the zoning is setup properly (assuming that is also changing) it should be very simple.

More important should be the discussion about getting updated to AIX 6.1 (at least). Although there may be VAR issues (do not support application X with AIX 6.1 - but then, are they supporting application X at all?)

By the way, you could simply connect to the new switch and type "cfgmgr", you will see the disks available through a new path and missing on the old path. Even no need to stop the database!