Kbase P165601: Process hung holding txe latch
| Autor |
Progress Software Corporation - Progress |
| Acesso |
Público |
| Publicação |
07/05/2010 |
|
Status: Unverified
SYMPTOM(s):
Process hung holding txe latch
Many processes blocked waiting for TXE latch.
protrace gathered from hung process indicates it is performing a semAdd operation
Database seems responsive to commands from promon
All hung processes are unresponsive.
Attempted to disconnect processes from database through promon.
Database log shows attempted disconnect of hung process but process does not go away or gain any CPU.
Excerpt of protrace:
semAdd
latWaitOnSem
latUsrWait
rlTXElock
tmend
startup parameters for -Mn and -Mpb do not add up correctly
FACT(s) (Environment):
All Supported Operating Systems
Progress 9.x
OpenEdge 10.x
CAUSE:
Improper configuration of -Mn and -Mpb for multiple login brokers contributed to shortage of semaphores.
FIX:
The parameters -n, -Mn, and -Ma are using internally to allocate resources from the operating system for control of structures within the database engine.
-n is the number of processes which will connect to the database (this includes all local and remote connections to the database)
-Mn is the total number of servers for the database
-Ma is the number of clients per server and is configured for each login broker
-Mpb is the number of server per broker and is configured for each login broker
-n should be large enough to encompass all self-service connections (including APWs, BIW, AIW, Watchdog, promon connections, online backups, etc.). It should also be configured large enough to account for all remote connections as well (-Mpb * -Ma for each login broker)
-Mn should be set based on this formula:
-Mn = sum( -Mpb + 1 for each login broker beyond the primary broker)