NetBSD-Users archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]

Re: dk wedges, or shooting oneself in the foot for profit and pleasure...

2009/11/11 Chavdar Ivanov <>:
> Hi,
> I had to replace an old and finally failed IBM Deathstar disk on one
> of my -current systems; there was this 120GB PATA disk lying around,
> looking too large for what was in use before, so I decided to make use
> of the versatility of GPT and wedges - which I had never come to try
> before - to split it in few slices with different usage and also iSCSI
> targets. Got my kernel (5.99.21 from yesterday) compiled with the all
> the DK* parameters turned on, booted single user, tweaked /etc/fstab
> and all was well. The problem was, my root disk is a reasonably fast
> SCSI 10K 18G one, three PATA disks are used for various slices (to
> include the new 120GB, so far unused). The SCSI disk gets its wedges
> auto-discovered last, so whenever I add a new one to the large disk I
> have to edit the fstab file. OK, that is in the GENERIC comments, so
> one is warned, of course. The trouble came when I made a wedge to use
> as an iSCSI target; it worked just fine, I unfortunately edited
> /etc/iscsi/targets specifying it and even started doing some data
> moving from a Vista initiator to get a feeling for the speed to
> expect. As I was about to leave the office, I decided to reboot and
> correct the fstab file with the new value of the root wedge, and so I
> did, forgetting to edit the targets file or to disable iscsi_target.
> The new root wedge name turned to be exactly the wedge which was
> created previously for the iscsi target... so my root slice was
> promptly offered as a target by the iSCSI initiator, the initiator
> still moving the data about... The end result is, fsck on the root
> wedge panics (I don't think the trace is useful, considering the
> obvious damage), so I will have to reinstall again (no big deal, as I
> did that a few weeks ago for different reasons, the data and the
> pkgsrc packages are intact).
> OK, we all know the ways users manage to shoot themselves in the feet
> are countless.
> The question: is there a way to force the wedge discovery to go in a
> predefined way to avoid the change of the root wedge every time, in my
> case to force the SCSI disk to be first in the queue?

Shot myself in the other foot today after a full with the DK*
enabled in GENERIC - the installation procedure cannot deal with
wedges completely - insisting to write to the underlying slices
instead. Well, live to learn... Going back to wedges only in GPT disks
I guess.

> --

Chavdar Ivanov

> ----
> Ted Turner  - "Sports is like a war without the killing." -

Stephen Leacock  - "I detest life-insurance agents: they always argue
that I shall some day die, which is not so." -

Home | Main Index | Thread Index | Old Index