Hello,
I'm using a Firebird 3 x86 Windows snapshot (SuperServer) from October
15, 2015 using Pavel's OLTP emulator to put load on Firebird, the trace
functionality and our FB TraceManager.
I'm using the following trace configuration:
database = oltpemul.fdb
{
enabled = true
log_statement_prepare = true
log_statement_free = true
log_statement_finish = true
log_procedure_finish = true
log_trigger_finish = true
print_plan = true
print_perf = true
log_function_finish = true
exclude_filter = %RDB$%
time_threshold = 0
max_sql_length = 2048
}
services
{
}
While the trace session is running everything looks ok, but when I'm
trying to stop the trace session (doesn't matter while the OLTP test is
still running or not) via a regular Services API call within FBTM, FBTM
locks up and a single thread in the firebird.exe process occupies a
single CPU core entirely.
The stack trace (not really helpful perhaps) of the offending thread
copied from ProcessExplorer is:
ntoskrnl.exe!KeWaitForMultipleObjects+0xc0a
ntoskrnl.exe!ExfReleasePushLockShared+0x91e
ntoskrnl.exe!KeWaitForSingleObject+0x19f
ntoskrnl.exe!PoStartNextPowerIrp+0xba0
ntoskrnl.exe!PoStartNextPowerIrp+0x1821
ntoskrnl.exe!PoStartNextPowerIrp+0x1a97
MSVCR100.dll!_time64+0x30
firebird.exe+0x797f8
MSVCR100.dll!_time64+0x30
Engine12.DLL!firebird_plugin+0x6f25d
firebird.exe+0x58d37
firebird.exe+0x5f011
firebird.exe+0x55fbb
firebird.exe+0x16390
firebird.exe+0x163ef
ntdll.dll!RtlAnsiCharToUnicodeChar+0xdf
Only a restart of fbserver.exe helps.
Running the same test scenario with Firebird 2.5.4 does not show the
same issue.
If you need any further information, please let me know.
Regards,
Thomas
------------------------------------------------------------------------------
Firebird-Devel mailing list, web interface at
https://lists.sourceforge.net/lists/listinfo/firebird-devel
