Philip,
I have a partial answer for you...
Chris Gerhard has put together a tool to trace SCSI commands.
This can be isolated to a single instance... see attaches "scsi.d".
Use this as a starting point for your testing, but focus the testing
to a single device instance.
Keep in mind that sd (SCSI Disk) and st (SCSI Tape) contain
overlapping & non-overlapping commands. See "http://t10.org"
for the standards... The top 3 standards to review are listed below.
The short answer is that SCSI Disks & SCSI Tapes look simular
from the Fibre Channel Exchange view.
SCSI-3 Primary Commands (SPC) or later: utilized by all SCSI devices.
SCSI-3 Block Commands (SBC) or later: utilized by most SCSI disks.
SCSI-3 Streaming Commands (SSC) or later: utilized by SCSI tapes.
Others may be utilized, but the above three is a good start.
Now as to "what is causing ... SCSI command", Sun Solaris[tm] will
conduct SCSI Standard INQUIRY, SCSI INQUIRY EVPD Page83, SCSI
INQUIRY EVPD Page80, TEST UNIT READY, and a couple others during
device discovery, Sun Cluster[tm] will conduct additional TEST UNIT READY
or WRITE(10) with zero (0) TransferLength to see the device is accessable,
and Sun Cluster[tm] will also conduct various RESERVE(6), RELEASE(6),
PERSISTENT RESERVE IN, & PERSISTENT RESERVE OUT commands
to manage SCSI-2/SCSI-3 Reservations. Other than that, you are normally
looking at application caused SCSI commands. So, depending on the
archiving software you are utilizing, there can be a number of SCSI
commands being utilized...
Good luck in your testing,
Joel Buckley,
Sun Storage Group
----- Original Message -----
From: Philip Brown <ppb at usc.edu>
Date: Friday, May 26, 2006 1:26 pm
Subject: [dtrace-discuss] getting all io to fibre tape drives
To: dtrace-discuss at opensolaris.org
> hi folks,
> was wondering if someone could help me with a quickie script.
we''re
> trying
> to get the equivalent of "snoop" for fibre devices :-) in
> particular, we''d
> like to find out what is causing any and all fibre packets to/from
> our fibre
> tape drives, from a particular machine.
>
> the sample "iosnoop.d" file, seems to be somewhat useful.
> However, it only seems to give raw "yes I''m doing
dev_strategy"
> calls. it
> doesnt tell me what actually triggered the call.
> (plus it gives ALL io, both sd and st. we only want to look at st)
>
>
> Apologies for being new to dtrace.. at this point, I know the
> basics, but I
> cant ramp up my knowlege fast enough, to tackle this Very Pressing
> production issue we are having, so I''m hoping someone on this list
> could
> help out and put something together before everyone vanishes for
> the weekend :-(
>
> _______________________________________________
> dtrace-discuss mailing list
> dtrace-discuss at opensolaris.org
>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: scsi.d
Type: application/octet-stream
Size: 15587 bytes
Desc: not available
URL:
<http://mail.opensolaris.org/pipermail/dtrace-discuss/attachments/20060526/dcde0c94/attachment.obj>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: Joel.Buckley.vcf
Type: text/x-vcard
Size: 348 bytes
Desc: Card for Joel Buckley <Joel.Buckley at Sun.COM>
URL:
<http://mail.opensolaris.org/pipermail/dtrace-discuss/attachments/20060526/dcde0c94/attachment.vcf>