[4918] in linux-scsi channel archive
Re: dual-attached disk
daemon@ATHENA.MIT.EDU (Santosh A Rao)
Fri Oct 16 11:51:46 1998
Date: Fri, 16 Oct 1998 10:47:33 +0530 (IST)
From: Santosh A Rao <santoshr@tagore.wipinfo.soft.net>
To: Andy Poling <andy@globalauctions.com>
Cc: Robert Dobozy <robo@idata.sk>, linux-scsi@vger.rutgers.edu
In-Reply-To: <Pine.LNX.3.96.981015000704.19200U-100000@roadrunner.realbig.com>
On Thu, 15 Oct 1998, Andy Poling wrote:
>On Wed, 14 Oct 1998, Robert Dobozy wrote:
>> > this. I don't have a lot of experience with using dual-hosted disks -
>> > the SCSI specification does allow for "reservations" so that one host can
>> > block others from using a device for some period of time, but I have no
>> > idea whether this is usable in practice.
>>
>> Yes it can be. If you want to share disk, but not in parallel. One machine
>> can reserve this disk and use it. Other machine will get reservation
>> conflict when accessing this device. When first machine is finished, it
>> must unreserve device.
>
>That could be a problem when the machine with the reservation is
>down/crashed...
the way "failover" systems are designed are such that a heartbeat
timer is mantained between the 2 initiators connected to the shared
dual/multi host disk. the heartbeat timer may involve a mechanism such
as an INQUIRY command issued periodically to the other initiator,
which would respond in target mode for the command. the driver for the
initiator would respond to the INQUIRY in one of the following ways :
i) the initiator would respond with normal INQUIRY data indicating
the primary initiator is alive.
ii) the INQUIRY to the initiator would time out indicating death of
the primary initiator.
iii) the initiator would respond with INQUIRY data with status
indicating a PanicInProgress, if the initiator has the OS capability
of determining the health of the system(i.e. determine if a panic is
currently in progress).
in cases (ii) and (iii) above, the secondary initiator would then
attempt to "take over" the shared multi/dual host disk. it would
usually do this by issuing a BUS DEVICE RESET or a BUS RESET which
results in the clearing of any previous reservation. thereafter, it
would proceed to RESERVE the disk and all I/Os to the disk would be
routed through the second initiator.
caution must be exercised when attempting such an activity since, if
the primary initiator is alive, it may result in the primary trying to
recover a lost reservation, which would leads to both initiators
commencing a struggle to take ownership of the disk.
- santosh
------------
>
>-Andy
>
>Global Auctions
>http://www.globalauctions.com
>
>
>-
>To unsubscribe from this list: send the line "unsubscribe linux-scsi" in
>the body of a message to majordomo@vger.rutgers.edu
>
-
To unsubscribe from this list: send the line "unsubscribe linux-scsi" in
the body of a message to majordomo@vger.rutgers.edu
