No relevant resource is found in the selected language.

This site uses cookies. By continuing to browse the site you are agreeing to our use of cookies. Read our privacy policy>Search


To have a better experience, please upgrade your IE browser.


Fake sessions appear because of internal queue overflow

Publication Date:  2014-07-18 Views:  123 Downloads:  0

Issue Description

 Quantity of session is doubled unexpectedly according SNMP monitoring system and at command line. Fake sessions don’t have IP address assigned and cannot be cut from command line.
of fake session:


<bras1>display access-user username JFK
  UserID  Username                Interface      IP address       MAC
  81877   JFK                     Eth-Trunk12.100  -                XXXX-XXXX-ce7b
  85275   JFK                     Eth-Trunk12.100  -                XXXX-XXXX-ce7b
  130504  JFK                     Eth-Trunk12.100  X.X.X.YYY        XXXX-XXXX-ce7b
  143098  JFK                     Eth-Trunk12.100  -                XXXX-XXXX-ce7b
  182054  JFK                     Eth-Trunk12.100  X.X.X.ZZZ        XXXX-XXXX-ce7b
  211567  JFK                     Eth-Trunk12.100  -                XXXX-XXXX-ce7b
  242098  JFK                     Eth-Trunk12.100  -                XXXX-XXXX-ce7b
  Total 7

Handling Process

Check from hidden view AAA queue status:
confirm that AAA internal software queue was overflowd at problem time:
 bras1-hidecmd]display message-queue 119

 Queue ID                   = 119
 Queue Name                 = AAA           
 Queue Mode                 = FIFO SYN
 Total Length               = 20480
 Current Length             = 0
 Used Percent               = 0%
 Peak Length                = 20480
 Peak Time                  = 2012.11.22 07:40:32
 Max Msg Size               = 16
Current version have software problem: when internal  queue is overflow, there is no protection and failure is occure at BRAS and Customer can see problem phenomena.

Root Cause

Root Cause is weak DNS module realization at ME60: if for some reason DNS is unavailable but ME60 need to resolve URL to IP address, then ME60 try three time to wait for DNS server reply and at this time can happens queue overflow, because many subscribers request DNS resolving and wait while ME60 finish this process. Also ME60 don't have any cache to keep frequently DNS name at hand, so even 100 subscribers request to resolve the same URL of web-portal, ME60 will issue 100 DNS requests to DNS server.
When ME60 will use DNS name resolving:
web-portal command line
web-portal attributes that send from RADIUS


 At version V1R6 use direct IP address notation instead of DNS name or upgrade to latest version, for example V6R5 is free from such problem