[Search for users] [Overall Top Noters] [List of all Conferences] [Download this site]

Conference smurf::ase

Title:ase
Moderator:SMURF::GROSSO
Created:Thu Jul 29 1993
Last Modified:Fri Jun 06 1997
Last Successful Update:Fri Jun 06 1997
Number of topics:2114
Total number of notes:7347

1939.0. "Frequency of Can't Reserve message?" by XFMV01::SHYAM () Tue Mar 11 1997 11:47

    
    
    Hello,
    
    What should be the frequency of 'Cant Reserve'
    			            ' SCSI Reservation failure' error in a
    normal ASE environment ?
    
    We get this error almost 90 % of the times while the services relocate 
    from one server to the other.
    
    ASE Version is 1.3 under Digital Unix 3.2D
    
    
    Thanks in advance
    
    
    Shyam
T.RTitleUserPersonal
Name		DateLines
1939.1Not 90%! NETRIX::"[email protected]"Gregory P. MyrdalWed Apr 02 1997 12:1811
Shyam,

I am not a scsi engineer, however, from my experience I would say there
might be something wrong with your scsi.  A normal relocation of a 
service should not produce scsi reservation failures.  We do see scsi 
reserve failures on reboot, but that's because ASE runs all of the stop 
scripts at boot time.  But normal relocations should work without these 
errors.

-- Greg
[Posted by WWW Notes gateway]
1939.2Nothing wrong, just a dumb applicationSMURF::KNIGHTFred KnightWed Apr 02 1997 17:5419
This type of error is typically triggered by an application.

Consider the following:

Host 1 reserves a device to offer a service.

Host 2 has an application that polls ALL scsi devices every
5 minutes to present some status information.  These polling
accesses will always FAIL when it gets to the device that
host 1 has reserved.

The trick is to find which application on host 2 is the one
that is doing the polling, and get it to stop polling the
device that is reserved by host 1.

Nothing wrong at all (in the H/W or S/W), just a non-ASE aware
application.

	Fred
1939.3Could NFS be the spoiler ?QCAV01::DEVARAJANMon Apr 14 1997 04:2223
    Re .1 & .2 :
    
    	Thanks for the info.  There are no SCSI reserve failures during a
    reboot.  This happens only when there is a shutdown of one of the
    members.  Actually there are 2 NFS services running on 2 AlphaServer
    2100s (one on each system).  Their device ids are /dev/rza8c and
    /dev/rzc8c.  On top of one of these services we have layered two Login
    services (as given in Internet AlphaSErver software).  The other NFS
    service does NOT have any "dependent" service on top of it.  Even
    then, we get SCSI reservation failures when this service (which does
    not have any dependents) relocates during a shutdown.
    
    	This happens most of the times.  As per Re .2, we can understand
    that NFS could be the spoiler in the case of the first service which
    has dependents.  But it does NOT explain why the second service also
    should have the same reservation problems (the one which does NOT have
    any dependents).
    
    	Any inputs or comments ?
    
    	Thanks
    
    PS : I work with Shyam (the original poster) in the same proj.