Home Page for the TeradataForum
 

Archives of the TeradataForum

Message Posted: Mon, 10 Mar 2003 @ 16:20:39 GMT


     
  <Prev Next>  
<<First
<Prev
Next>
Last>>
 


Subj:   One possible cause of 3111's
 
From:   Anomy Anom

<-- Anonymously Posted: Monday, March 10, 2003 09:14 -->

Fresh off the presses:

DR67117 - Don't time out transaction when running out of AWT's

CUSTOMER-PROBLEM-DESCRIPTION:

Transactions are getting timed out with error 3111 when the system is running out of AMP worker task (AWT).

FAULT-ANALYSIS DETAILS:

The deadlock detection logic mistakenly interprets that the transactions have request queued up at the mail box waiting for available AMP worker tasks to be deadlocked. And since these transactions do not have legitimate reason to be blocked, they will be aborted by the dispatcher and cause 3111 time out errors.

This error only happened on a system when all of the AMP worker tasks are busy and there is only one available AMP worker tasks left. At this time, if the Dispatcher sends a transaction with an AMP step doing row redistribution to the mailbox, the only available AWT will become a sender task. However, since it is the last available AWT, the receiver does not get spawned, and the sender gets stuck waiting for the receiver. This transaction does not get CPU time, it also not being blocked and subjected to being aborted by the Dispatcher.



     
  <Prev Next>  
<<First
<Prev
Next>
Last>>
 
 
 
 
 
 
 
 
 
  
  Top Home Privacy Feedback  
 
 
Copyright for the TeradataForum (TDATA-L), Manta BlueSky    
Copyright 2016 - All Rights Reserved    
Last Modified: 27 Dec 2016