I`ve run some heavy query on auxiliary (NON-production) database that is placed on same machine as our production. This query was running too long and I decided to kill it:
SQL>delete from mon$attachments; quit;
Immediatelly after it we detected that no more new connections to production can be established via TCP. Connections to production via local protocol were avaliable.
The firebird.log does NOT contains any interesting info - only messages with errno=-104.
I could create backtrace of fb_smp_server process when we could not to connect (see attach #1) and after restart linux and restoring common workload of out users (see attach #2).
Unfortunatelly I could NOT make log of fb_lock_print -d /var/db/firebird/production.fdb -a -m -c >fb_lock_print_dead.txt because log file has been created with zero size.
SQL> show version;
ISQL Version: LI-V188.8.131.52448 Firebird 2.5 // SUPERCLASSIC
Firebird/linux AMD64 (access method), version "LI-V184.108.40.206448 Firebird 2.5"
on disk structure version 11.2
SQL> show database;
Number of DB pages allocated = 1888000
Sweep interval = 0
Forced Writes are OFF
Transaction - oldest = 218095653
Transaction - oldest active = 218307431
Transaction - oldest snapshot = 218307431
Transaction - Next = 218365577
ODS = 11.2
Default Character set: WIN1251