[nv-l] snmpcollect deferrals on NetView for Windows

[James Shanks <jshanks@us.ibm.com>]

Those of you who are experiencing deferred collections might benefit from the text of APAR PJ29256, which added more error checking and tracing into snmpcollect for NetView for Windows 7.1.4, so that it is easier to identify the reason for your deferrals.   However, what it says about why deferrals occur and what you can do to minimize them applies to all earlier versions as well.   Hope this helps.

****************************************************************                
* USERS AFFECTED: All NetView for  Windows snmpcollect users.  *                
****************************************************************                
* PROBLEM DESCRIPTION: NetView for Windows snmpcollect often   *                
*                      defers collections for one hour when    *                
*                      the MIB browser says all is OK, and     *                
*                      the reason is not obvious.              *                
****************************************************************                
* RECOMMENDATION:                                              *                
****************************************************************                
A collection is deffered, by default for one hour, when                         
the node does not respond to an snmpget.  This deferral time                    
can be changed using "snmpcollect -D deferminutes" in                           
snmpcollect's startup, but additional tracing has shown in the                  
lab that the collections are deferred because snmpcollect                       
sends out an SNMP request but never receives a reply.                           
                                                                                
The collection that has no reply is the reason that ALL the                     
collections for that node are deferred. Normally, the trace                     
will show a "read select timeout failure" and then the                          
collections getting deferred. The problem is the that the                       
reply is never received. It is not coming in late. The                          
question now is, "Who  is at fault?"                                            
  It seems that sending a large burst of requests in a small                    
time interval could be a problem somewhere in the network                       
even though it seems to the user that it shouldn't be.                          
                                                                                
snmpCollect's trace normally will not have any errors logged                    
other than the "read select timeout failure", so additional                     
tracing would be helpful to isolate the problem.                                
NOTE: If you're seeing deferred collections, even though the MIB                
Browser works with them, it probably is because the snmpcollect                 
daemon isn't waiting long enough for the MIB reply. The default                 
of 3 seconds is sometimes not enough time.                                      
To allow  more time for collections, do the following:                          
1. Go to Options > SNMP > select the object on which you're                     
seeing the deferred collections                                                 
2. Select Properties                                                            
3. Look at the "Retry SNMP Requests Every __ Seconds for a                      
Maximum of __ Tries" fields                                                     
If these fields are blank, enter values like 5 seconds and 3                    
retries.                                                                        
If they are not blank, increase the values in the fields.                       
PROBLEM CONCLUSION:                                                             
 . . . [ removed the part about how to turn on new tracing
         in snmpcollect for NetView 7.1.4   ]          
Collection Deferrals are normal occurances from time                            
to time when network problems occur. If the customer does                       
not want the collections deferred for an hour then                              
he should change this using "snmpcollect -D Deferminutes".                      
  Also it is helpful to stagger the collection times. It was                    
found in the lab during testing that adding a small                             
delay after each send request, which slows down the burst of                    
sends, did cause fewer collections to get deferred.                             
                                                                                

James Shanks
Level 3 Support  for Tivoli NetView for UNIX and Windows
Tivoli Software / IBM Software Group