Also, possibly related to this (appears on same production servers) - sometimes 
all fb_inet_server.exe processes stop any disk i/o activity and wait for one 
process, which eats 100% of one CPU core. 

 Stack trace for hung thread look like this:
 fb_inet_server!down_grade+0x145
 fb_inet_server!down_grade+0x156

 

 ....tens of recursive down_grade calls
 

 fb_inet_server!down_grade+0x156

 fb_inet_server!blocking_ast_bdb+0x69
 fb_inet_server!Jrd::LockManager::blocking_action+0x170
 fb_inet_server!Jrd::LockManager::blocking_action_thread+0x11f
 fb_inet_server!Jrd::LockManager::blocking_action_thread+0x9
 fb_inet_server!`anonymous namespace'::threadStart+0x55
 msvcr80!_endthreadex+0x3b
 msvcr80!_endthreadex+0xc7
 kernel32!BaseThreadInitThunk+0xe
 ntdll_77770000!__RtlUserThreadStart+0x23
 ntdll_77770000!_RtlUserThreadStart+0x1b
 

 


 I has debug dump and lock print for hung process.
 
 


Reply via email to