Conference noted::pwv50ift

4296.0. "Incomplete PW Install Problems???" by GRANPA::BARABIA () Mon May 19 1997 10:38

    A couple weeks ago one of our system admin folks tried to install
    PATHWORKS on a new node we are bringing into our existing cluster.
    Our current cluster consists of 2 7000 series Alphas (nodes MINNIE and
    DAISY). They share a common system disk and are at VMS V6.2. We've
    another Alpha 8400 (node BAMBI) that has its own system disk and its at
    VMS V6.2-1H3. We've applied a number of VMS patches as well. PATHOWRKS
    is V5.0D ECO3. Our new node is another 8400 (PLUTO) with its own system
    disk at V6.2-1H3. 
    
    The person who attempted the install of PATHWORKS on PLUTO started the
    install during normal working hours, thinking since PLUTO had its own
    system disk he could do the install without affecting the other active
    cluster members. He unfortunately did not realize that the PATHWORKS
    install needed to shutdown PAHTWORKS on the other cluster members.
    Attempting to re-start PATHWORKS was not successful since the other
    members required re-booting before PATHWORKS could be re-started.
    Why is that neccessary? I'm confused why the install on the 4th cluster
    member warrants a re-boot of all the other PATHWORKS Cluster members?
    MINNIE and DAISY were not re-booted for 2 days after the install. BAMBI
    was not re-booted till a week later. Could this cause us problems?
    
    About a week after the install. they needed to shut down PLUTO, the
    newly installed system. The VMS Shutdown or PLUTO hung and they forced
    a crash on PLUTO. We then experienced a suspended PWRK$LMSRV process on
    node MINNIE which alos impacted PATHWORKS access on all cluster. We
    were forced to then re-boot MINNIE. MINNIE re-booted fine and PATHWORKS
    access was ok on the cluster. An analysis of the system dump file
    inidciated and outstanding lock could not be resolved when PLUTO tried
    to dismount a disk device that had locks taken out on by the LMSRV
    process on node MINNIE. Unfortunately the system dump file was not
    large enough and they could not get a better look at what the LMSRV
    process was doing. Could we run into some sort of locking problem
    caused by the incomplete PATHWORKS install on PLUTO? We also found out
    that when PLUTO hung trying to shutdown, PATHWORKS was not active on it
    however we subsequently did another VMS shutdown on PLUTO with
    PATHWORKS active and PLUTO shutdown normally. When the tried another
    VMS shutdown of PLUTO with PATHWORKS not active and we experienced the
    same hung VMS shutdown.
    
    My thought is that somehow the incomplete install of PATHWORKS may be
    causing problems. All of this is documented with the CSC under log
    numbers C970429-122 and C970508-3859, but they are scratching their
    heads also.
    
    I know this is kind of vague, but thats all I have for now and of
    course the customer wants answers.
    
    Any input would be appreciated.
    
    Thanks,
    
    Barry A
    

I can only give you some answers for the install questions:

>    The person who attempted the install of PATHWORKS on PLUTO started the
>    install during normal working hours, thinking since PLUTO had its own
>    system disk he could do the install without affecting the other active
>    cluster members. He unfortunately did not realize that the PATHWORKS
>    install needed to shutdown PAHTWORKS on the other cluster members.

I'm not sure why the installer didn't realize this.  The PATHWORKS install
procedure explicitly warns the installer that PATHWORKS must be shut down
on the other cluster members, and gives the installer the option to exit
out of the install procedure if the installer doesn't want to do this.

>    Attempting to re-start PATHWORKS was not successful since the other
>    members required re-booting before PATHWORKS could be re-started.
>    Why is that neccessary? I'm confused why the install on the 4th cluster
>    member warrants a re-boot of all the other PATHWORKS Cluster members?
>    MINNIE and DAISY were not re-booted for 2 days after the install. BAMBI
>    was not re-booted till a week later. Could this cause us problems?

You and I know that the new system has a separate system disk, but the
PATHWORKS install procedure doesn't.  Technically, separate system disks
is unsupported, except for mixed VAX and Alpha clusters, and then only
for one system disk for each architecture type.  Rebooting several days
after the install should cause no problems.