Conference noted::pwv50ift

4247.0. "V5.0E ECO1 - PCOMS: failed to allocate NET LOGON message buffer" by LNZALI::BACHNER (Mouse not found. Click OK to continue) Fri Apr 04 1997 11:30

Cluster with OpenVMS/VAX V7.0
Pathworks V5.0E ECO1
two boot servers act as member servers in a domain with an WinNT 3.51 PDC

Since we upgraded from V5.0E to ECO1, the servers seem to be irresponsive to new
connection requests. NETLOGON is not running and cannot be started manually
(NET2187: The service control is busy.)

We've found the following error logged in PWRK$LMLOGS:PWRK$LMMCP_node.LOG over
and over again:

 4-APR-1997 15:01:03.75 23E0084E:003379B0 PCOMS: failed to allocate NET LOGON
message buffer
 4-APR-1997 15:01:03.80 23E0084E:003379B0 PCOMS: error occured at source line
1987
 4-APR-1997 15:01:03.82 23E0084E:003379B0       free message buffers  = 90
 4-APR-1997 15:01:03.86 23E0084E:003379B0       free logon messages   = 0
 4-APR-1997 15:01:03.86 23E0084E:003379B0       free process elements = 6 (1
message buffers)
 4-APR-1997 15:01:03.87 23E0084E:003379B0       free fork elements    = 31 (1
message buffers)
 4-APR-1997 15:01:03.87 23E0084E:003379B0       free fd elements      = 6 (2
message buffers)
 4-APR-1997 15:01:03.87 23E0084E:003379B0       free pipe elements    = 5 (2
message buffers)
 4-APR-1997 15:01:03.87 23E0084E:003379B0 PCOMS: cannot get message buffer for
send

The log files show that this problem started after the installation of ECO 1.

I tried the workaround suggested when this problem was reported for earlier
versions of Pathworks (added a [PCOMS] section to PWRK.INI to increase the
maximum number of IPC and NETLOGON messages). At first, it did not seem to help,
until PATHWORKS was restarted on *both* servers.

Today a customer reported the same problem.

Is there a reason why the default parameters are not sufficient ?  On our
inhouse system, PATHWORKS load ist *very* light (at an average less than 5
sessions per server; maybe a bit more because of the 'nameless' sessions from
the NT 4.0 workstations and servers).

Thanks for listening,
Hans.

    To solve this we need a system crash on the node which logged the pcoms
    error messages.
    The reason for this is  that pcoms is using shared memory to
    communicate between the PATHWORKS processes.
    We do not need a system dump of the other node.
    
    Rob.

    Found a cause.
    
    Disable the alerter service.
    
    Fix will be in v50f
    No patch kit because workaround is sufficient.
    
    Adrie

Thanks for your response !

Hans.

    I too appreciate the response :)

    there could be other causes. check out the response to a prev. note
    which explains how why when where
    (and howmuch should be credited to my bank account - but the question
    is by whom?)
    
    Adrie