Subject | FB 1.5.2 terminates abnormally (-1) |
---|---|
Author | precari0 |
Post date | 2005-05-16T17:04:09Z |
Hello everyone,
Recently I upgraded a customer site from firebird 1.0.3 to 1.5.2 for
performance reasons.
Since then, fb crashes every 2 or 3 days.
Here's the environment and some more info:
- Windows 2000 Server (with all current updates)
- 1GB RAM
- P4 CPU (several MHz :) , hyperthreading disabled.
- One 100Mbit ethernet
- Windows' TCP KeepAliveTime is set to 10min
- DB file is ~400MB (grows ~3MB/day), page size is 4KB, cache size is
2048
pages
- Firebird memory usage is consistently between 50 and 90MB (seen in
taskmanager). No leaks evident.
- RemoteAuxPort is set to 3051
- "gstat -h" was being run every 30 seconds and its results recorded.
There
is no transaction leak.
- DB runs 24x7
- About 120 connections during peak usage.
- At least 40 connections during night shift.
- Use a couple of very simple UDFs (Delphi).
- Heavy use of events. Every connection registers interest in at
least 3
different events (some register up to 10).
- The number of different events "posted" from psql is ~100. There
are a
couple of events being posted at least once every second (automated).
- Every client is running on gds32.dll 1.5.2
- Every client is running on msvcrt and msvcp60 distributed with fb
1.5.2
- If I replace fb 1.5.2 with 1.0.3 everything runs fine (but much
slower).
Do you guys have _any_ idea of what may be causing fb1.5.2's
termination?
If the crash was related to UDFs I'd see it in the log files, right?
Anyway, I can post the source and declaration if necessary.
Anything else I should check? Known gotchas? More info?
Thanks for you input,
Douglas Heron Tosi
PCNET Informática ltda
CTVoicer Computer Telephony ltda
mailto : douglas em ctvoicer . com . br
41 21039300
41 99219950
Recently I upgraded a customer site from firebird 1.0.3 to 1.5.2 for
performance reasons.
Since then, fb crashes every 2 or 3 days.
Here's the environment and some more info:
- Windows 2000 Server (with all current updates)
- 1GB RAM
- P4 CPU (several MHz :) , hyperthreading disabled.
- One 100Mbit ethernet
- Windows' TCP KeepAliveTime is set to 10min
- DB file is ~400MB (grows ~3MB/day), page size is 4KB, cache size is
2048
pages
- Firebird memory usage is consistently between 50 and 90MB (seen in
taskmanager). No leaks evident.
- RemoteAuxPort is set to 3051
- "gstat -h" was being run every 30 seconds and its results recorded.
There
is no transaction leak.
- DB runs 24x7
- About 120 connections during peak usage.
- At least 40 connections during night shift.
- Use a couple of very simple UDFs (Delphi).
- Heavy use of events. Every connection registers interest in at
least 3
different events (some register up to 10).
- The number of different events "posted" from psql is ~100. There
are a
couple of events being posted at least once every second (automated).
- Every client is running on gds32.dll 1.5.2
- Every client is running on msvcrt and msvcp60 distributed with fb
1.5.2
- If I replace fb 1.5.2 with 1.0.3 everything runs fine (but much
slower).
Do you guys have _any_ idea of what may be causing fb1.5.2's
termination?
If the crash was related to UDFs I'd see it in the log files, right?
Anyway, I can post the source and declaration if necessary.
Anything else I should check? Known gotchas? More info?
Thanks for you input,
Douglas Heron Tosi
PCNET Informática ltda
CTVoicer Computer Telephony ltda
mailto : douglas em ctvoicer . com . br
41 21039300
41 99219950