Thanks, I tried firing up Gregor Kobler's FBMonitor (similar to the old but dead Sinatica Monitor).
I found that OAT is 222085103 while OIT and OST (what's that?) is 261140768 and growing. I assume the OAT must have got stuck somehow quite a while back, since it's so far behind the others. The machine was rebooted for upgrades rather recently, and I can't see any stuck transactions, so the OAT must be a zombie one. No auto sweep (interval set to 0), so I'm trying a manual sweep now. Guess I should schedule that weekly or something. I couldn't see anything interesting happening with the trans numbers when I exited the batch app. Will restart it after the manual sweep and see if anything interesting shows up. The slowdown, if still there, would probably happen after I go to sleep, so I'll have to follow up tomorrow. Other info about the database, in case it says anything: Page size: 16384 ODS: 12.0 Page buffers: 614400 (roughly 9.5 Gbyte) Sweep interval: 0 (like I mentioned) Forced writes: off (yes, I know the risks) From databases.conf: DefaultDbCachePages = 600K (same as Page buffers above, of course) FileSystemCacheThreshold = 1000K From firebird.conf: FileSystemCacheSize = 40 TempCacheLimit = 4G MaxUnflushedWrites = 100 MaxUnflushedWriteTime = 60 ServerMode = Super All other performance params default. Some auth and access restriction params are non default, but I assume that's irrelevant. System has 40 Gbyte RAM and the temp disks have 70+ Gbyte free. Regards, Kjell Den 2020-04-29 kl. 14:00, skrev Karol Bieniaszewski [email protected] [firebird-support]: > Hi > > There are many possibilities without access i can only hint you: > > Look at MON$Tranasctions maybe you have active one which stop garbage > collecion. > > Look also at sort buffer setting if firebird.conf > > Look at settings about buffers in database itself (gfix -h show you > value). > > Look also at automatic sweep settings (also gfix -h show you value). > > Update Firebird to most recent version official is 3.0.5 but i use > most recent snapshot without problems. > > More can be tell after some details provided. > > regards, > > Karol Bieniaszewski > > [Non-text portions of this message have been removed]
