Severe Performance Degradation of Firebird Server – Multiple Instances and Strange Subprocesses

[TechAnton]

Hello everyone,

We have encountered a significant performance issue with our Firebird server. A query that previously completed in 3 hours failed to finish even after 48 hours. Upon investigation, we found that there were four active Firebird instances running on the server, none of which were heavily utilizing CPU or RAM. Additionally, suspicious subprocesses were observed, including those related to Windows Update.

The issue was resolved by terminating all Firebird processes through Task Manager and performing a full server reboot, after which performance returned to normal.

My question is: is such behavior typical for Firebird? Could having multiple instances or these external subprocesses impact the server’s performance?

I have attached a screenshot of the Task Manager for your reference.

Thank you in advance for your assistance!

[PedroCostinha]

Hi! I wouldn't consider a query taking 48+ hours instead of 3 hours to be normal Firebird behavior, but the presence of multiple Firebird processes does not necessarily indicate a problem.

The first thing I'd check is the server architecture: Classic, SuperClassic, and SuperServer can have different process and resource usage patterns.

Since a reboot restored performance, I would investigate what changed while the problem was happening:

1. Check firebird.log and Windows Event Viewer for errors, restarts, disk warnings, or backup activity.
2. Check disk latency and queue length, not just CPU and RAM usage. A process can be blocked on I/O while using very little CPU.
3. Inspect active connections and transactions through Firebird's monitoring tables (MON$ATTACHMENTS, MON$TRANSACTIONS, and MON$STATEMENTS).
4. Compare the execution plan and record statistics of the slow query before and after the problem, if available.
5. Check whether Windows Update, antivirus scanning, or backup software was accessing the database files or saturating the storage device.

I would avoid terminating all Firebird processes as a routine fix, since that can interrupt active transactions.

Could you share the Firebird version, server architecture, database size, storage type, and whether the multiple processes were all associated with the same Firebird installation? That would help distinguish expected process behavior from an actual performance issue.