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

Conference decwet::networker

Title:NetWorker
Notice:kits - 12-14, problem reporting - 41.*, basics 1-100
Moderator:DECWET::RANDALL.com::lenox
Created:Thu Oct 10 1996
Last Modified:Fri Jun 06 1997
Last Successful Update:Fri Jun 06 1997
Number of topics:750
Total number of notes:3361

360.0. "NSRD cores when adding clients" by CUSTOM::STAFFORD () Wed Jan 29 1997 06:59

    
    	I am using a ALPHA 2100 running DUNIX 3.2G and NSR server
    	4.2a. The clients are a mix of NT Alpha and Intel running
    	WNT Server 3.51 sp5 with 4.3 NSR on Intel and 4.2a NSR on ALPHA.
    	
        We have two servers with about 70 clients each of the above
        configuration.
    
    	The problem I am having is the NSR Server keeps crashing when
    	I add or modify a NT client. This does not happen every time
    	but often enough to make it a pain.
    
    	When adding the client the server sends out the NSR ping but
    	does not get a complete packet back. The server starts parsing 
    	the message, recognizes the incomplete reply and cores.
    
    	Here is an extract of the messages in daemon.log
    
    	nsr_ping: buf returned was type: NSR client desfcription;
    	pools supported: Yes;
    	remote user: SYSTEM;
    	groups: Administrators, Everyone;
    	arch: Windows NT;
    	clienttest: res_parsebuf returned Unexpected end of input
    	clienttest: cpus attribute not found for client
    
    	After than NSRD is no longer running and we have to restart NSR
    	on the sever.
    
    	We are surprised that the server recognizes that the clients
        response was inclomplete and even logs a message but then decides
    	to core. I would thing the server would handle incomplete messages
    	and keep going on its way. This crashing is causing lots of
    	concerns since now we have to be very careful about adding clients
    	or modifying client definitions in fear that it might ping the
    	client and crash the server
     
    
    	We have two UNIX NSR servers, one I upgraded to NSR 4.2b to see
        if that would help. It did not.
    	
    	IS there a patch for the server to have it stop crashing?
    
    	Thanks
    	Brian	
T.RTitleUserPersonal
Name
DateLines
360.1is this for a customer, or Digital Internal??DECWET::EVANSBe a Point Of Light!Wed Jan 29 1997 11:273
obviously this is a bug, and needs to be fixed in our next release (at 
 a minimum) and patched if possible.

360.2For a customerCUSTOM::STAFFORDWed Jan 29 1997 13:1814
    
    
    	This is for a customer..
    
    	If there is any chance of getting a patch that would be great.
    	As I said we have two Digital NSR servers here but the customer
    	also has a SUN NSR server. Of course he compares urs to Suns
    	in areas of reliability etc.
    
    	Anything that can be done to resolve this problem would help
    	the effort in the field.
    
    	Thanks
    	Brian
360.3understood,and...DECWET::EVANSBe a Point Of Light!Wed Jan 29 1997 17:375
just so you know where we are right now... 

  20 IPMT cases (1/2 CLDs, 1/2 SPRs), and a rapid development cycle
 for the next release. We'll sincerely do what we can. Stay tuned to this
 note thread for news. It may be a few days at least....
360.4Looking into it...DECWET::FARLEEInsufficient Virtual um...er....Fri Jan 31 1997 12:1413
I am looking into this area of code for the next release, so I'll
dive into this problem as well.  It just crashed my test server...

There seem to be two problems:
1) The NT client is not always returning the complete buffer of 
	attributes to the UNIX server during the "ping".

2) The UNIX server is not dealing with the incomplete buffer in a 
	graceful way.

I can solve #2, but I may or may not be able to solve #1.

Kevin
360.5DECWET::KOWALSKITime's not for savingThu Feb 06 1997 09:1721
Kevin and I have completed investigations on this
problem and are beginning the process of issuing 
patches for both the Digital Unix server and
the V4.3 NT Alpha client.  A patch for V4.3 
NT Intel client will have to come from Legato.

In summary:

- at client definition time, the 4.3 NT clients
  are responding to the Digital UNIX server's request
  for client attributes with information the
  server can't parse.  This occurs even when
  the client timing problem (below) does not
  occur.

- the 4.3 NT clients have a timing problem when
  responding to this attribute query which 
  occasionally results in incomplete information
  being returned to the Digital UNIX server.

Mark