two filer root volumes on the same loop?

[joeycitrix]

hello

question

im redoing my loop because im upgrading from fas3050c to fas3140 and planning to put all FC disks on one loop and all SATA disks on second and third loop.

currently, the two filers root volume is on SATA disks. will this affect my loop configuration if both of my root volume is on the same loop. the primary loop path is on filerB and im going to do MPHA. or do i need to copy one root volume to FC disks for this to work.

thanks

[scottgelb]

Along with loop considerations, the 3050 may not have software disk ownership enabled, but the 3140 requires software disk ownership.  If you type "disk show -v" on the 3050 and there is no output, then you have hardware ownership and will have to boot to maintenance mode (ctrl-c for special boot menu, then option 5) then at the maintenance prompt on each node type "disk upgrade_ownership".  Make sure to follow the ugprade advisor and match the flash card code on the 3140 or netboot to make it match (3050 uses "e" elf code and 3140 uses "q" code).  I usually upgrade the 3050 elf code to match the flash on the 3140 prior to the swap (using the e code)...then the 3140 boots just fine off flash then push the q code again so it replaces the e code on disk...and follow the instructions for disk reassign and destroy mailbox local/partner... then /etc/rc and hosts for any interface changes with the head swap.

Having both root aggregates on the same loop isn't my preference in case of a shelf failure..and ontap should give an informational message about this configuration....but you said you have 1 fc loop and 2 sata loops... with mpha, I would spread my root aggregate across both sata loops.  You can move the drives with the system halted, then it will boot no problem with the drives moved.  Assuming a 3 drive root aggregate with 2 loops, put 1 of the root drives on one loop then the other 2 on the other loop and then if a loop fails, you stil have your root aggregate available.

for example:  with the two loops...move the 3 root disks so you have root across both loops and with disk assignment it doesn't matter if a mix in the loop/shelf to each 3140 node.

loop2  node1-aggr0-rg0-disk1

loop3  node1-aggr0-rg0-disks2-3

loop2  node2-aggr0-rg0-disks1-2

loop3  node2-aggr0-rg0-disks3

[joeycitrix]

scott

my fas3050c has been changed to SDO a few months back.

would my config works like this

filerA: loop 1- 6 FC disk shelves

filerB: loop 1- old filerA root aggr

filerB: loop 2- old filerB root aggr

if the configuration is what i want to do, how does FilerA knows which root aggr to use and the same for FilerB

thanks

[scottgelb]

Assuming you are keeping 3 separate loops...but adding 2 new paths to each loop (for mpha), then I would just move 1 drive from each node's root aggregate to the other shelf on a separate loop.  FilerA and FilerB know their disks regardless of which loop they are on.  "disk show -v" will show disk assignment of each individual disk drive.  A single disk shelf can have disks for both nodes in the same shelf/same loop.  If you halt both nodes and shuffle the drives in any order (within their drive type) the system will boot and find all its disks).

My assumption is you are keeping 3 loops with 4 connecdtions each (12 fas connections, 6 per node)... then have your root aggregate available between 0b/1b and 0c/1c loops in the example below.

FC ports on each controller

0a/1a                                     fc loop

0b/1b                                     sata loop1 (mix of disks from both nodes)

0c/1c                                     sata loop2  (mix of disks from both nodes)

[joeycitrix]

scott

so FilerA is intelligent enough, if all shelves are online(filerB still power off), it sees two root volume on its loops and will pick only one root volume to boot? Or is this a manual process in maintenance mode to point FilerA to its correct root volume and once that completes, power up FilerB and it should only see one root volume and automatically know that the root volume belongs to it.

[aborzenkov]

Every filer sees only those disks that have been explicitly assigned to it (disk assign). On those disks exactly one root volume exists.

It may happen that if you add disks from another system and assign them they contain foreign root volume. I do not know how filer handles it and which one will be selected during boot; when this happened once on customer system foreign root volume was marked as "next root" meaning filer would boot from it next time; we quickly destroyed it as it was not needed

[scottgelb]

With software disk ownwership filerA will only use its own disks assigned to it.  filerB disks will show as partner disks.  The only exception is if there is a cluster takeover event by filerA and disk reservations bring up the partner disks as the partner filerB under filerA.  You will need to disk reassign from the 3050 systemid to the 3140 systemid for each A/B node as part of the controller swap.  When A boots with B down, A will find its disks and boot and B will find its disks and boot.

Since you are not reassigning another systems disks (only existing) you won't have foreign aggregates/volumes.  If you had a foreign aggr/vol that showed up as "diskroot" that would be odd...diskroot becomes root on next boot.  You have to "vol options volname root" to mark a volume as diskroot...but it is possible the foreign volume was marked as root before it was connected to a foreign system and kept the designation.  In your case there should not be any foreign volumes if you are only doing a controller swap on existing disks.