scsictl immediate_report

[Peter Kovacs 1.0rc]

Hi,

I am testing an rx2620 with two scsi drives. I would like to have both report write operations immediately. Do I have to configure both disks separately, or is this done on the controller level?

I assumed that I have to call scsictl separately for both disks, but I have already trouble determining which disk is mapped to which device.

ioscan gives the following:

bash-3.00# /sbin/ioscan -fun -C disk
Class I H/W Path Driver S/W State H/W Type Description
============================================================================
disk 0 0/0/2/0.0.0.0 sdisk CLAIMED DEVICE TEAC DV-28E-N
/dev/dsk/c0t0d0 /dev/rdsk/c0t0d0
disk 1 0/1/1/0.0.0 sdisk CLAIMED DEVICE HP 73.4GST373454LC
/dev/dsk/c2t0d0 /dev/rdsk/c2t0d0
disk 2 0/1/1/0.1.0 sdisk CLAIMED DEVICE HP 73.4GST373454LC
/dev/dsk/c2t1d0 /dev/dsk/c2t1d0s2 /dev/rdsk/c2t1d0 /dev/rdsk/c2t1d0s2
/dev/dsk/c2t1d0s1 /dev/dsk/c2t1d0s3 /dev/rdsk/c2t1d0s1 /dev/rdsk/c2t1d0s3

I'd think /dev/rdsk/c2t0d0 represents one of the disks, but:

bash-3.00# scsictl -m immediate_report /dev/rdsk/c2t0d0
immediate_report = 0
bash-3.00# scsictl -m immediate_report=1 /dev/rdsk/c2t0d0
bash-3.00# scsictl -m immediate_report /dev/rdsk/c2t0d0
immediate_report = 0

Any help appreciated.

Peter

[Hein van den Heuvel]

Hello again Peter,

scsictl would work in the individual drive level. For your 2620 there is no controller (like a smartarray) but just a 'pass through' adapter. No value added there.

The disks in questions MIGHT not support this bit, but I would be surprised. Maybe I'll try it on my box at home tonite.

From the man page:

"WARNINGS
Not all devices support all mode parameters and commands listed above. Changing a mode parameter may have no effect on such a device."

Best I can tell your disks would be /dev/rdsk/c2t0d0
and /dev/rdsk/c2t1d0
The c2t1d0sX are probable partitions for the EFI support? Just guessing.

The SATA drives which you mention in the other topic and which perform better really open you up to significant risks. They have no battery backup with their cache, and may loose a lot of data if somethign goes wrong. Also, best I know SATA drives tend to have a lower reliability. Together not a good combo for a serious production box, but great for a quick hack or to show amazing benchmark numbers.

fwiw,
Hein.

[Peter Kovacs 1.0rc]

Hein,

Thank you for your quick reply.

The command seems to work with the other disk:

bash-3.00# scsictl -m immediate_report /dev/rdsk/c2t1d0
immediate_report = 0
bash-3.00# scsictl -m immediate_report=1 /dev/rdsk/c2t1d0
bash-3.00# scsictl -m immediate_report /dev/rdsk/c2t1d0
immediate_report = 1

The ioscan output shows the two disks to be of the same model (ST373454LC: Seagate Cheetah 15K.4 73GB Hard Drive). Could it be possible that support for this feature varies for individuals of the same model?

Thank you for warning me on the hazards of write caching. But you guessed right (again): it is about benchmarks . (Being a tiny software development company, we have just one "production machine", the one on which our website and CVS repository is running. And transactional consistency is not an issue there.)

Peter

[Denver Osborn]

Maybe it's related to disk firmware? Have you compared the f/w revs of both drives?

diskinfo -v /dev/rdsk/c2t1d0
diskinfo -v /dev/rdsk/c2t0d0

-denver

[rick jones · Accepted solution]

Just a WAG, but are there mounted filesystems on both discs?


ISTR there being some sort of ir kernel tunable default_ir perhaps?

[Hein van den Heuvel]

I think Rick has it, or at least close.

Your c2t0d0 is probably the mounted boot drive and the system trying to prevent you from shooting yourself in the foot.

My system behaves the same.

I did try mounting an lvol on the 3rd drive (the second drive is an OpenVMS boot disk) and scsictl can change the mode bit before and after mount.

Full session log attached.
Summary:
/dev/rdsk/c2t0d0 immediate_report = 0
/dev/rdsk/c2t1d0 immediate_report = 0
/dev/rdsk/c3t2d0 immediate_report = 0
>scsictl -m immediate_report=1 /dev/rdsk/c2t0d0
>scsictl -m immediate_report=1 /dev/rdsk/c2t1d0
>scsictl -m immediate_report=1 /dev/rdsk/c3t2d0
>scsictl -m immediate_report /dev/rdsk/c2t0d0
immediate_report = 0
>scsictl -m immediate_report /dev/rdsk/c2t1d0
immediate_report = 1
>scsictl -m immediate_report /dev/rdsk/c3t2d0
immediate_report = 1

I highly recommend investing in a third drive!
You have the slot.
You have an extra channel.
Nice for tests. Much more freedom.

For my first tests I did not have an extra drive with the right packaging. I used an older drive from a HP/Compaq wintel server. It came in that larger (1/2" higher, 1" deeper, long lever) aluminum finned package you might have seen. I popped the drive itself, as it has had the right connector on the back. Used a wood wedge ( 1/2 a cloth pin :-) to keep it in place. Good enough for some tests.

Regards,
Hein.

[Peter Kovacs 1.0rc]

Thank you for all of your replies.

Denver,
"rev level" is the same for both disks: "HPC2" Is it the f/w version? Or is the f/w version contained in "Additional inquiry bytes"? The only differing item in the diskinfo output is "Additional inquiry bytes".

Rick, Hein,
There are no mounted filesystems on /dev/dsk/c2t1d0, just raw logical volumes:

bash-3.00# pvdisplay /dev/dsk/c2t1d0
--- Physical volumes ---
PV Name /dev/dsk/c2t1d0
VG Name /dev/vg_oraraw
PV Status available
Allocatable yes
VGDA 2
Cur LV 12
PE Size (Mbytes) 4
Total PE 17499
Free PE 14341
Allocated PE 3158
Stale PE 0
IO Timeout (Seconds) default
Autoswitch On

And /dev/vg_oraraw is a raw volume group.

I think there are filesystems on /dev/dsk/c2t0d0. However, I seem to be unable to verify this:

bash-3.00# pvdisplay /dev/dsk/c2t0d0
pvdisplay: Couldn't find the volume group to which
physical volume "/dev/dsk/c2t0d0" belongs.
pvdisplay: Cannot display physical volume "/dev/dsk/c2t0d0".

SAM is also confusing.

The "Disk Devices" list shows:

0/0/2/0.0.0.0 1 Unused -- 0 DVD-ROM Drive
0/1/1/0.0.0 1 Unused -- 70007 HP 73.4GST373454L
0/1/1/0.1.0 1 LVM vg_orara 70007 HP 73.4GST373454L

The "Volume Groups" list shows:
vg00 LVM 0 of 69088 1 9
vg_oraraw LVM 57364 of 69996 1 12

And it is vg00 which I suspect is on /dev/dsk/c2t0d0.

The machine was shipped with "vg00" containing the system pre-installed. This volume group is missing from the "Disk Devices" list. Is this normal???

"vg_oraraw" has been added by my precious self. Did I, perhaps, commit some fatal errors when creating it? (I do not remember erasing anything in the process. I still have the first commands in my shell history:

/sbin/pvcreate /dev/rdsk/c2t1d0
/sbin/vgcreate /dev/vg_oraraw /dev/dsk/c2t1d0

so chances are that I did not touch /dev/dsk/c2t0d0 along the way.)

Thanks a lot!

Peter