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

Conference taec::temip

Title:Telecom. Management Information Platform conference
Moderator:TAEC::LAVILLAT
Created:Fri Mar 08 1991
Last Modified:Fri Jun 06 1997
Last Successful Update:Fri Jun 06 1997
Number of topics:725
Total number of notes:2431

722.0. "TNS Server Problem in ASE" by TAEC::STANARD (Bob Stanard - MCI/Mission Critical- CSC/Colorado) Mon May 26 1997 15:39

SOFTWARE VERSION: 

   TFRBASEV310     installed       TeMIP Framework Base System (level 1 rev E)
   TFRSERVERV310   installed       TeMIP Framework Server (level 1 rev E)
   TFRTNSCLRKV310  installed       TeMIP Name Service Clerk (level 1 rev E)     
   TFRTNSSRVV310   installed       TeMIP Name Service Server (level 1 rev E)
   

   Digital Unix 3.2g

   ASE 1.3

CONFIGURATION:

   2 node ASE (failed on 1000, 4000, 2100s)
   TeMIP and TNS Server are installed as a common service in the ASE.
   TNS Clerk is installed on each node

SYMPTOM:

   TeMIP and TNS operate normally until the active server that contains the TNS
   service fails. The action scripts function normally, and the service is 
   apparently failed over to the backup server. Attempts to access TNS 
   following failover, however, result in the clerk being unable to communicate
   with the TNS server:

     #   tnscp "dir dir *"
         DIRECTORY
         DIRECTORY  some_ns:.*
                                       AT  09-MAY-1997:15:41:01
     Error on entity: some_ns:.
     Unable to communicate with any TNS server 
     Function: tnsEnumChild

PROBLEM:
   The ASE software does not register it's use of IP ports that are less than
   1024 in /etc/services. Although TNS has registered port 1000 in
   /etc/services, the ASE software allocates this port and blocks incoming
   requests to the TNS server.

WORKAROUND:

   You must change the ports that TNS uses in /etc/services:

   o Stop the tns processes on the server : tns_stop
   o Modify the port number for the TNS service in the /etc/services file
   o Start the tns server : tns_start

   As the port number is in the TNS server advertisment message, it is not
   necessary to specify this port number on the client machines.

SOLUTION:

   The problem has been reported to Digital Unix Engineering. It is expected
   that Trucluster ASE will properly reserve the ports it uses.

T.RTitleUserPersonal
Name
DateLines
722.1Similar ProblemNNTPD::"[email protected]"winkler::COLA1Tue May 27 1997 13:1346
SOFTWARE VERSION: 

   TFRBASEV310     installed       TeMIP Framework Base System (level 1 rev E)
   TFRSERVERV310   installed       TeMIP Framework Server (level 1 rev E)
   TFRTNSCLRKV310  installed       TeMIP Name Service Clerk (level 1 rev E)   
 
   TFRTNSSRVV310   installed       TeMIP Name Service Server (level 1 rev E)
   + latest Patch 001 for all subsets

   Digital Unix 3.2g

   ASE 1.3

CONFIGURATION:

   2 node ASE (4100)
   TeMIP and TNS Server are installed as a common service in the ASE.
   TNS Clerk is installed on each node


We have currently a similar problem with more or less the same configuration 
(TEMIP and TNS Server) configured as a common service.

On our platform TeMIP and the TNS server can be relocated to the
Failover-Server
but the TNS clerk will run extremly slowly on the Failover-Machine for 
subsequent operations. 

E.g. show domain *  -> 60 seconds for only two registered domains
     tnscp show .temip.<director_name> all attr -> 10 seconds to wait

     The clerk does not take advantage of the cache data for more than one
     attempt (confidence level= medium)



For all other cases (no Failover) one can recognise acceptable runtimes. 
Maybe our configuration is slightly different, cause we configured the TNS
clerk
on the Failover machine explicitely by use of tns_clerk_setup before starting 
the temip_setup -s ... procedure.

We are currently working on the problem ....


[Posted by WWW Notes gateway]
722.2Customer solution implemented for problem similar to .0NNTPD::&quot;[email protected]&quot;martinMon Jun 02 1997 13:1322
FROM - DECnotes Reply 2090.3 in Conference ase on smurf




``The Customer has implemented following solution.'' by
NNTPD::"[email protected]" (martin +
customer) on Fri, 30 May 1997 08:26:56 EDT

CUSTOMER SOLUTION::

In /etc/rc.conf they have set cms_conf to off, and they have not seen any
malfunctions. They would offcourse like to know if this is a problem , doing
so, Other than they do not have the same type of logging  fram the trace.d 
and submon.

So things are running fine for them and their testing at DSC Communications

regards MArtin


[Posted by WWW Notes gateway]