I have an Ubuntu 10 Server LTS 64 bit box running Firebird Classic server v2.1.3.

I am having an issue with it in that the Mutex wait % statistic from FB_Lock_Print is very high, currently around 42% which I know is way way too high.

What I am struggling with is why this is the case. There are two hard drives for the OS in a mirrored configuration, with a 200Gb Kingston SSD for the major databases.

All disks are showing in Ubuntu as connected at SATA 6Gbps and the load isn’t that high, currently only 16 connections. I have tweaked the firebird.conf parameters as recommended by IBSurgeon with no real improvement. The major load on this server is synchronisation traffic between databases, we have a process that transfers data from one database to the other by exporting a linked set of data from one database and importing it into the other database.

The 2 main databases are on the SSD and they are 31Gb & 45Gb in size. There are some additional databases on the hard drives but they are only 12Gb & 13Gb.

If I run disk speed tests using hdparm –tT /dev/sdx, I get results that are consistent from one run to the next and are consistent with the type of drive being tested (HDD or SSD).

I have an almost identical server running similar loads except that it doesn’t have the SSD and the Mutex wait % on that one is only 4% or so, right where it should be.

Does anyone have any ideas as to what might be causing the high mutex wait %?

Cheers,

Neil Pickles - neil@...