On Thu, May 31, 2001 at 09:55:36AM -0700, W. Curtis Preston wrote:
I think it should #2.
1) Some Unix flavors doesn't support reserve/release
feature for example SGI. Not sure about NT/2000
2) Solaris supports this feature.
But.... The "right" description for DLT tape drives we received from
Veritas:
Q-DLT7000 = 1,0x38,0,0x3D639,4,0x82,0x83,0x84,0x85,2;
# man st
NAME
st - driver for SCSI tape devices
..............skiped...
<options> is a bit pattern representing the devices, as
defined in /usr/include/sys/scsi/targets/stdef.h. Valid
flags for tape configuration are:
ST_VARIABLE 0x0001
ST_QIC 0x0002
ST_REEL 0x0004
ST_BSF 0x0008
ST_BSR 0x0010
ST_LONG_ERASE 0x0020
ST_AUTODEN_OVERRIDE 0x0040
ST_NOBUF 0x0080
ST_KNOWS_EOD 0x0200
ST_UNLOADABLE 0x0400
ST_SOFT_ERROR_REPORTING 0x0800
ST_LONG_TIMEOUTS 0x1000
ST_BUFFERED_WRITES 0x4000
ST_NO_RECSIZE_LIMIT 0x8000
ST_MODE_SEL_COMP 0x10000
ST_NO_RESERVE_RELEASE 0x20000
ST_READ_IGNORE_ILI 0x40000
ST_READ_IGNORE_EOFS 0x80000
ST_SHORT_FILEMARKS 0x100000
ST_EJECT_TAPE_ON_CHANGER_FAILURE 0x200000
ST_RETRY_ON_RECOVERED_DEFERRED_ERROR 0x400000
So, 0x3D639 has ST_NO_RESERVE_RELEASE bit set to 1
ST_NO_RESERVE_RELEASE
The ST_NO_RESERVE_RELEASE flag disables the use of
reserve on open and release on close. If an
attempt to use a ioctl of MTRESERVE or MTRELEASE
on a drive with this flag set, it will return an
error of ENOTTY (inappropriate ioctl for device).
> I'm getting conflicting information from a couple of different sources, and
> I KNOW that someone out there knows the real answer.
>
> How does SSO (for NetBackup & BE if possible) managed contention for the
> robot? It would seem that this could be done in one of two ways:
>
> 1. It uses the SCSI Reserve command when it starts to use the arm.
> If no other host is using it, the command will work, and it will proceed.
> If another host is using it, the command will fail, and it will have to
> be retried.
>
> 2. It doesn't need to use the SCSI Reserve command, because it is accepting
> all the requests.
> Requests for the robot are "queued," and each host needing the robot is
> told to wait it's turn.
> SCSI Reserve is not needed, because the software will insure that only
> one robotic command is issued at a time.
>
> 3. Same as #2, except it also uses SCSI Reserve. It doesn't NEED to, but
> it does anyway.
> Why? I don't know. ;)
>
> It would SEEM that #2 would be preferable.
>
>
>
> ---
> W. Curtis Preston
> Principal Consultant for Storage Designs, your storage experts
> Webmaster: http://www.backupcentral.com Phone: 760 631 7991
>
> _______________________________________________
> Veritas-bu maillist - Veritas-bu AT mailman.eng.auburn DOT edu
> http://mailman.eng.auburn.edu/mailman/listinfo/veritas-bu
--
----
Andrew Shinkarev
Network Systems Specialist
Abbott Laboratories
(847) 9387559
shinkara AT pprd.abbott DOT com
|