Conference decwet::networker

Your server setup looks fine, as does the saveset specification.

Look at the client setup with View Details on.  Check the value 
for the client parallelism parameter.  It should be greater than 
1.

If the client parallelism parameter is greater than 1, then try 
the command "savegroup -pnv <group>" and post the results here.

Wes

    
    
    Hi
    
    I posting the result of your request 
    
    garfield : Nsr Server(ALPHA)
    aladdin  : nsr client(HP)
    
    
    [garfield:/]:# savegroup -pnv testgrp-2
    aladdin:/                                 level=incr
    aladdin:/www                              level=incr
    aladdin:/home                             level=incr
    aladdin:/dolphins                         level=incr
    aladdin:/tmp                              level=incr
    aladdin:/oracle/CRISIS                    level=incr
    aladdin:/oracle/SYST                      level=incr
     2/19/97 11:16:09 savegroup: Run up to 4 clients in parallel
     2/19/97 11:16:09 savegroup: aladdin:probe                             
    started
    savefs -s garfield -g testgrp-2 -p -n -l full -R -v -F / /www /home
    /dolphins /
     2/19/97 11:16:30 savegroup: aladdin:probe succeeded.
    rcmd aladdin, user root: `savefs -s garfield -g testgrp-2 -p -n -l full
    -R -v -'
    asavegrp: authtype nsrexec
    type: NSR client description;
    pools supported: Yes;
    arch: HP 9000;
    CPU type: 9000/897;
    CPUs: 2;
    IP address: 12.26.3.10;
    MB used: 3090;
    OS: HP-UX B.10.01;
    version: 4.2.3;
    save set: path=/, arg=/, level=full,diskno=0,
    max_sessions=1,stype=save,\
    path=/www, arg=/www, level=full, diskno=0, max_sessions=1, stype=save,\
    path=/home, arg=/home, level=full, diskno=0, max_sessions=1,
    stype=save,\
    path=/dolphins, arg=/dolphins, level=full, diskno=0, max_sessions=1,
    stype=save\
    path=/tmp, arg=/tmp, level=full, diskno=0, max_sessions=1, stype=save,\
    path=/home/ftp, arg=/home/ftp, level=full, diskno=0, max_sessions=1,
    stype=save\
    path=/oracle/CRISIS, arg=/oracle/CRISIS, level=full, diskno=0,
    max_sessions=1, \
    path=/oracle/SYST, arg=/oracle/SYST, level=full, diskno=0,
    max_sessions=1, styp;
    parallelism: 8
    --- Probe Summary ---
                 
    aladdin:/                             level=full, dn=0, mx=1,
    vers=pools, p=8
    aladdin:/              level=full, save as of Thu Jan  1 00:00:00
    GMT+0000 1970
    aladdin:/www                          level=full, dn=0, mx=1,
    vers=pools, p=8
    aladdin:/www           level=full, save as of Thu Jan  1 00:00:00
    GMT+0000 1970
    aladdin:/home                         level=full, dn=0, mx=1,
    vers=pools, p=8
    aladdin:/home          level=full, save as of Thu Jan  1 00:00:00
    GMT+0000 1970
    aladdin:/dolphins                     level=full, dn=0, mx=1,
    vers=pools, p=8
    aladdin:/dolphins      level=full, save as of Thu Jan  1 00:00:00
    GMT+0000 1970
    aladdin:/tmp                          level=full, dn=0, mx=1,
    vers=pools, p=8
    aladdin:/tmp           level=full, save as of Thu Jan  1 00:00:00
    GMT+0000 1970
    aladdin:/home/ftp                     level=full, dn=0, mx=1,
    vers=pools, p=8
    aladdin:/home/ftp      level=full, save as of Thu Jan  1 00:00:00
    GMT+0000 1970
    aladdin:/oracle/CRISIS                level=full, dn=0, mx=1,
    vers=pools, p=8
    aladdin:/oracle/CRISIS level=full, save as of Thu Jan  1 00:00:00
    GMT+0000 1970
    aladdin:/oracle/SYST                  level=full, dn=0, mx=1,
    vers=pools, p=8
    aladdin:/oracle/SYST   level=full, save as of Thu Jan  1 00:00:00
    GMT+0000 1970
    aladdin:index                        level=full, dn=-1, mx=0,
    vers=pools, p=8
    aladdin:index          level=full, save as of Thu Jan  1 00:00:00
    GMT+0000 1970
    
    
    
    
    
    
    
    
    
    
    
    
    
    
    
    
    
    
    
    

The result of the savegrp tells me that the HP client is reporting
that all of its filesystems are on the same disk.

(look in the savegrp -pnv results; Each of the filesystems reports
"diskno = 0".)  

Savegrp takes this information and attempts to NOT start more than one session
at a time on the same physical disk, to avoid thrashing the disk.

It appears that there is a bug in the HP client such that it is reporting
all filesystems as being on the same physical disk.

    
    
    Hi
    
    According to the  note of 417.3, there is no action just say result.
    Note of 417.3 means  filesystems mounted on the same disk can,t backup
    parallel. If that is true, I don,t belive.
    To confirm  your reply,  as following i test.
    
    filesystem:
               /dev/rz1a   /
               /dev/rz1g   /usr
               /dev/rz0c   /users
               /proc
    
    
    Client Save Set : /, /usr,/users, /proc
    Parallelism : 4
    active device : 4
    session per device : 1
    
    CLient : Alpha DU4.0
    Server : ALpha DU4.0
    
    Result:
    [garfield:/]:# savegroup -pnv DEFAULT
    garfield:/                                level=incr
    garfield:/usr                             level=incr
    garfield:/proc                            level=incr
    garfield:/users                           level=incr
     2/21/97 13:50:05 savegroup: Run up to 2 clients in parallel
     2/21/97 13:50:05 savegroup: garfield:probe                            
    started
    savefs -s garfield -g DEFAULT -p -n -l full -R -v -F / /usr /proc
    /users
     2/21/97 13:50:11 savegroup: garfield:probe succeeded.
    rcmd localhost, user root: `savefs -s garfield -g DEFAULT -p -n -l full
    -R -v -'
    asavegrp: authtype nsrexec
    type: NSR client description;
    pools supported: Yes;
    arch: decaxp;
    CPU type: alpha;
    CPUs: 1;
    hosts: 1;
    IP address: 12.26.2.114;
    kernel arch: OSF1;
    License units: 1050;
    NAS Client: No;
    OS: OSF1 V4.0;
    version: V4.2A;
    save set: path=/, arg=/, level=full, diskno=1, max_sessions=1,
    stype=save,\
    path=/usr, arg=/usr, level=full, diskno=1, max_sessions=1, stype=save,\
    path=/proc, arg=/proc, level=full, diskno=1, max_sessions=1,
    stype=save,\
    path=/users, arg=/users, level=full, diskno=0, max_sessions=1,
    stype=save ;
    parallelism: 8
    --- Probe Summary ---
    
    garfield:/                            level=full, dn=1, mx=1,
    vers=pools, p=8
    garfield:/             level=full, save as of Thu Jan  1 00:00:00
    GMT+0000 1970
    garfield:/usr                         level=full, dn=1, mx=1,
    vers=pools, p=8
    garfield:/usr          level=full, save as of Thu Jan  1 00:00:00
    GMT+0000 1970
    garfield:/proc                        level=full, dn=1, mx=1,
    vers=pools, p=8
    garfield:/proc         level=full, save as of Thu Jan  1 00:00:00
    GMT+0000 1970
    garfield:/users                       level=full, dn=0, mx=1,
    vers=pools, p=8
    garfield:/users        level=full, save as of Thu Jan  1 00:00:00
    GMT+0000 1970
    garfield:index                       level=full, dn=-1, mx=0,
    vers=pools, p=8
    garfield:index         level=full, save as of Thu Jan  1 00:00:00
    GMT+0000 1970
    
    
    Actually, we are very urgent. If this test is going well, they will buy
    TL810(10ea). But we don,t have no solution yet. 
    
    Give a information what we must to do.
    
    thanks
    
    
     
    
    
    
    
    
    
      
     
    
    

	I'm not sure what you are asking in .4

	The / and /usr partions are on one disk and the /users is on
	an other, the /proc is not backed up since it does not have
	real data.  A backup of this system should start two sessions,
	one of / or /usr and the other with /users.  The probe data
	shows this by giving different disk numbers to the partions
	on different drives.  The HP client did not.  All that is needed
	for everything to work is for the HP client to report its drives
	properly.

One way to work around this is to have multiple client 
definitions, each of which specifies a one (or a few) filesystems
in its saveset field.  Try assigning these clients to a single 
group and see what happens.

You may need to assign the client definitions to separate groups
(keep the number of groups small, otherwise other problems will 
arise).  The groups could then be launched at the same time and
the backups should proceed in parallel. 

Another way is to run multiple saves on the client system using a 
cron job.



The Alpha client (garfield) is reporting two different disks. 
The HP client (aladdin) is not.  I assume the customer is running 
V4.2 of the HP client.

From .2

 save set: path=/, arg=/, level=full,diskno=0, max_sessions=1,stype=save,\
 path=/www, arg=/www, level=full, diskno=0, max_sessions=1, stype=save,\
 path=/home, arg=/home, level=full, diskno=0, max_sessions=1, stype=save,\
 path=/dolphins, arg=/dolphins, level=full, diskno=0, max_sessions=1, stype=save\
 path=/tmp, arg=/tmp, level=full, diskno=0, max_sessions=1, stype=save,\
 path=/home/ftp, arg=/home/ftp, level=full, diskno=0, max_sessions=1, stype=save\
 path=/oracle/CRISIS, arg=/oracle/CRISIS, level=full, diskno=0, max_sessions=1, \
 path=/oracle/SYST, arg=/oracle/SYST, level=full, diskno=0, max_sessions=1, styp;

The HP client is indicating that the filesystems /, /www, /home, 
/dolphins, /tmp, /home/ftp, /oracle/CRISIS, /oracle/SYST are all 
on diskno=0.

Is the HP system using an advanced file system (JFS?) or a volume 
manager?  If so, which ones?  There is a similar problem logged 
with Legato (IC 211336, 960730.03), and I'd like to make sure 
this is the same issue.

Regards,

Wes

.4 tried to duplicate the test on an AXP - that *does not* have the
 bug in save/savefs that HP has.

.3 explained what savefs was doing accurately, please read it carefully,
 and do the same type of tests on the AXP and you'll see that savefs
 has diskno=? where ?=0,1,2,3 -- on the HP, ?=0,0,0,0...

this is why you claim .3 is ambiguous. This is incorrect.

Did this customer buy the HP clientkit from Digital, or from Legato?

Let me attempt to be more clear:

You have uncovered what appears to be a bug in the HP client.

The symptoms of the bug are that it reports all filesystems to be
on the same physical disk.

We cannot fix HP clients.  Legato will have to do that.  We can work the
issue with them if you wish, but we are not capable of fixing it.

savegroup is designed to not start more than one save on the same physical
disk concurrently.  You can, however, manually start multiple saves
at the same time.  This may be a temporary workaround.  You will not,
however, be able to use savegroup to automatically schedule the saves, and
have them go in parallel.

Is this more clear to you?

Kevin