HP DL385G1 Smart Array 6i AMD64 FBSD 6.3

[Todorov]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Ulf Zimmermann написа:
| On Sun, May 04, 2008 at 01:11:05AM +0300, Todorov wrote:
|> -----BEGIN PGP SIGNED MESSAGE-----
|> Hash: SHA1
|>
|> Ulf Zimmermann ??????:
|> | On Sun, May 04, 2008 at 12:11:54AM +0300, Todorov wrote:
|> |> |>> Backup, yes do it. But here is how to do it:
|> |> |>>
|> |> |>> hpacucli ctrl slot=0 ld 1 add drives=allunassigned
|> |> |>>
|> |> |>> This should add the additional drives to your first logical
drive and
|> |> |>> expand it. Whenever I do this, I reboot now so FreeBSD sees the
|> |> |>> new larger physical disk.
|> |> |>>
|> |> |>> In most cases now you have to update the partition table, either
|> |> |>> by addition another slice or changing the size of your FreeBSD
slice.
|> |> |>>
|> |> |>> If the file system you want to grow is the last one in
disklabel, you
|> |> |>> can use disklabel to change the size of it, also the total
line (c:)
|> |> |>> needs to grow.
|> |> |>>
|> |> |>> And then finally you can use growfs on the file system.
|> |> |> Growfs on 6.X will destroy your UFS2, and mabye UFS1, filesystem.
|> |> |> The patch in bin/115174 appeared to fix growfs but I recently had
|> |> |> a file system related panic on a machine with a growfs expanded
|> |> |> filesystem.
|> |> |>
|> |> |>   reboot after panic: ffs_alloccg: map corrupted
|> |> |>
|> |> |> That panic may be unrelated to growfs but that machine is the only
|> |> |> 6.3 machine I have in production with a grown file system.  The
|> |> |> 800G fileystem had a few percent used when it paniced.  It was
being
|> |> |> brought into service as a Cyrus IMAP replica.
|> |> |>
|> |> |> A working growfs was a requirement for me to move from rock solid
|> |> |> 4.11.  It would really be nice if the filesystem guru's could take
|> |> |> a look at the growfs problem.
|> |> |>
|> |> |> John Capo
|> |> |> Tuffmail.com
|> |> | I just recently did a growfs in the describe method on:
|> |> |
|> |> | log01 ulf /home/ulf > uname -a
|> |> | FreeBSD log01.autc.com 6.1-STABLE FreeBSD 6.1-STABLE #0: Tue Jul 18
|> |> | 06:32:28 PDT 2006     root@:/usr/obj/usr/src/sys/LOG  i386
|> |> |
|> |> | That machine had a pair of 72GB drives, I swapped one drive at a
|> time to
|> |> | 300GB. Then ran "hpacucli ctrl sl=0 ld=1 modify size=max", rebooted
|> into
|> |> | linux (small partition on the server) to use the fdisk there,
rebooted
|> |> | back into FreeBSD, it saw the larger drive, larger partition. Then I
|> |> | edited "disklabel -e /dev/da0s4" to edit slice c: and f:. Ran
growfs,
|> |> fsck.
|> |> | and voila I had more space for my logs.
|> |> |
|> |>
|> |> I've checked today and what was before - 2 disks in Logical RAID 1+0
|> |> drive. By default they were mirrored as they are not 4. Today I've
added
|> |> two more (4 total) and make EXPAND to the logical drive. This process
|> |> finished but the BIOS tool and the FBSD see the size of one disk, not
|> |> two as it should be.
|> |>
|> |> 4x146GB in RAID 1+0, FBSD see 136GB as /dev/da0 (shown in the message
|> |> during boot). Also the BIOS says logical drive is 136G and is ok.
|> |>
|> |> Should do EXTEND of the logical drive also - is this an option?
|> |
|> | hpacucli ctrl sl=0 array A show
|> | hpacucli ctrl sl=0 ld 1 show
|> |
|> | That will show the current status. If you are going to just create
a new
|> | partition/file system anyways, you don't have to extent, you could just
|> | create a new logical disk, it would show as da1. To do that:
|> |
|> | hpacucli ctrl sl=0 array A create type=ld raid=1
|> |
|> | That will use whatever space is available in array A (if the current
|> status
|> | is that the two new drives have been added to array A).
|> |
|> | Of in case the drives are still seperate you can create an array B.
|> |
|>
|> Let me ask since I got confused...
|>
|> 4 disks in RAID 1+0 (let's assume I don't want to extend the FS) should
|> be seen as two RAID 1 's ?
|>
|> Should I see the RAID 0 somewhere also?
|>
|> Can you please provide some output of the dmesg or camcontrol inquiry .
|
| What does "hpacucli ctrl sl=0 array A show" show?
|

It worked :)

I couldn't use hpacucli under 6.3 AMD64 FreeBSD (if anyone could please
let me know, as it didn't recognize the controller). So I used the
SmartStart CD.

I did expansion of the current RAID 1+0 (RAID1 actually as there were
only 2 disks) with another 2 disks. So the expansion finished in about 6
hours and I saw OK for the volume BUT the size was the same as the RAID1
- - 146GB (4x146GB disks inside the logical drive). So I started asking
myself if the process did something so I went onsite next day and
started the ACU again - there was an option to EXTEND the SIZE as there
was another 146GB free space unassigned to the previous size. I did this
(few seconds) and the size became 280GB (MAX).

Currently I have 280GB - there was a problem with the fdisk - it was
warning me for the slices not starting on the beginning of the track
which led to non bootable system (data was there but the boot loader
told me - wrong part info). So I did it again and then override the boot
loader - this time it was ok.

In the FBSD I see /dev/da0 as RAID 1 which is strange but nevermind.

Thanks Ulf for encouraging me to test this!
I had backup for the other scenario but I did it w/o reinstall of the OS
which was nice.

There were some messages in syslog:

May  4 15:42:13 opteron kernel: ciss0: *** Parity/consistency
initialization complete, logical drive 0
May  4 15:42:13 opteron kernel: ciss0: logical drive 0 () completed
consistency initialisation

which I had to monitor for anycase if they are regular or just something
with the last restart.

Regards,
Todorov
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.9 (FreeBSD)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iEYEARECAAYFAkgeJa8ACgkQibJkIG65HMeiDwCg9w/DS6IZgyiFByPLxhXyGVG1
in4AoMh2zVUHHMqZNMJrMf0ds3zGVLKq
=lB3C
-----END PGP SIGNATURE-----