From: <az...@cs...> - 2003-09-26 14:50:13
|
Hello! > > > DefaultDbCachePages = 2048 # Expecting not related to Lock manager > ^^^^^^^^^^^^^ Buffer cache is very related to lock manager. Each > cached buffer consumes at least 100 bytes of lock manager memory. I was totally worng :-( > > > LockMemSize = 1048576 > ^^^^^^^^^^^^^ But this amount should be more than enough to cache > 2x2048 pages (~= 400k lock manager memory). Strange. > The only reasons of failure in this case I see are either you do a lot > of work and your buffer cache grows much above 2048 pages or there is > a lot of lock contention (that is unlikely). Only two major things were done before error occured: select from table with ~30k rows to display some initial overview (returned about 150 rows) same select with join to adresses table restrirted to specified region (~ 10 rows) -- Lock mem error occured. The second select use distintc whereas the first not. Well there are some small select before that to load settings (tables about 100 rows or less). But because I receive this error also from MiTec Interbase Performane Monitor (only connection test, no sense to run it on CS) it sounds to me to be independent of running SQL. > Try to increase lock manager memory until the problem disappear or > revert to default size of buffer cache. It would cost some time - i don't have this environment near to me. > Win32 CS automatic lock manager memory pool resizing is disabled in > RC5 and RC6 builds. It will be re-enabled in Firebird 1.5 final > release and is already enabled in Firebird 2.0. May be automatic resize could help. Thanks for the good explanation! Ales Zeleny ----------------------------------------------------------------------- Ales Zeleny (OK1UUE) ----------------------------------------------------------------------- Due to technical difficulties tomorrow has been postponed indefinitely. |