Kbase P21222: lock stack underflow error appears during database shutdown
| Autor |
Progress Software Corporation - Progress |
| Acesso |
Público |
| Publicação |
3/14/2003 |
|
Status: Unverified
FACT(s) (Environment):
AIX 4.3.3
FACT(s) (Environment):
Progress 9.1D
SYMPTOM(s):
lock stack underflow error appears during database shutdown
protrace from lock stack underflow error shows:
latlatch
ebsdfreu
ebsdsthp
ebsdstds
796 errors appear with errno 32:
Error writing msg, socket=<n> errno=<n> usernum=<n> disconnected. (796)
Connection times to database increase continually.
promon > User Control indicate resource waits - LKT, BKT, BUFF.
CHANGE:
Upgraded from v8 to v9.
CAUSE:
/usr/include/sys/errno.h would suggest that errno 32 is:
#define SIGPROF 32 /* profiling time alarm (see setitimer) */
SIGPROF indicates that the virtual timer has expired. The signal is delivered to a process when a virtual timer alarm that that process has scheduled with the setitimer system call expires. The default behaviour is that the process gets terminated.
The time limit might expire because the process is forced to wait for a resource.
Taken together - the 796 error with this errno, the occurrence of the lock stack underflow and the associated protrace suggest resource contention. The promon output can confirm this.
If -spin isn't set and no writers are started, they would be the first thing to look at.
FIX:
1. Increase -spin.
2. Ensure that writers are run.