Not very sure, but there may be a bug still around when a shelf is replaced on a head.if you remove a shelf and replace it with another but with a different serial number you may have an issue that only a reboot will fix.

On Jun 1, 2010 6:19 PM, "Steve Losen" <scl@sasha.acc.virginia.edu> wrote:


Hello toasters,

We need to replace a disk shelf on a FAS3050 clustered pair and I
was wondering if I could shut down just the filer that owns the
shelf and leave the partner up, even though the partner is
connected to the shelf.

We had a fan fault on a PSU, replaced it, and later on the fan
fault came back.  Replaced it again and the same thing happened,
so now we must replace the shelf.  It is in the middle of a
shelf chain and I must disconnect it and physically remove it from
the cabinet.

On the FAS3050 disk ownership is hardware assigned.  Filer "A"
owns all of the shelves and disks on its 0a FC adapter and the
bad shelf is on this chain.  The partner filer "B" is also
connected to these shelves using its 0b FC adapter and the ESH2
"B" modules on the shelves.

I am wondering if I can leave filer "B" up.  I was thinking along
these lines:

cf disable

shut down filer "A"

power off bad shelf, disconnect, and remove

install replacement shelf, move disks, PSUs and ESH2 modules
from old shelf to replacement shelf.

set shelf id

connect and power on

wait for shelf to boot and disks to spin up

boot filer "A"

cf enable

I am sure that filer "B" will be unhappy because its 0b loop will
be broken, but I hope it will stay up because all of the disks
that B owns will remain available.  Only partner disks owned by "A"
will disappear.

Will filer "B" return to normal when its 0b loop is reconnected?

We are running ONTAP  7.3.1.1

I know it is safer to shut down both filers, but the resulting
downtime would be more than twice as disruptive.

Steve Losen   scl@virginia.edu    phone: 434-924-0640

University of Virginia               ITC Unix Support