HP-UX: Session Hanging Indefinitely With Event "kfk: async disk IO"
(Doc ID 2246456.1)
Last updated on AUGUST 04, 2018
Applies to:Oracle Database - Enterprise Edition - Version 22.214.171.124 to 126.96.36.199 [Release 11.2 to 12.2]
Sessions hang for many hours or indefinitely until it got killed manually.
The hang occurs randomly with wait event "kfk: async disk IO" on all 6 nodes of the RAC database without any particular pattern. The same hang occurs in single instance environment with ASM.
- Based on DB dump traces, Async I/Os are passed to OS and they are stuck. ASM is hung waiting for I/O completion for more than a day (1705 min 24 sec)
- Based on OSWatcher metrics on I/O, the disk is continuously accessed throughout the hang period. However some I/Os are stuck at OS level.
- At session level the hang is seen on same wait event 'kfk: async disk IO' stuck indefinitely until the report job is killed manually with "alter system kill session..command"
- Async I/O pending on UFS: which is why 'kfk: async disk IO' hung. Trace shows there are read or write I/Os pending below ASM causing sessions to hang.
- Hang is only at single session / process and there is no blockers or hang chains.
- No errors are observed on OS or storage level.
- ASM configuration is in the I/O calls during process cleanup and stuck on pending I/O's.
- Truss of hung process shows "/dev/async" closed but session continues to read and reap pending I/O's.
- Instance Alert log has error (ORA-00041: active time limit exceeded - session terminated) when session hangs.
Short stack dump:
ksedsts()+592<-ksdxfstk()+48<-ksdxcb()+3376<-sspuser()+688 <-<kernel<-_read_sys()+48<-_read()+224<-skgfospo()+608<-$cold_skgfrwat()+208 <-ksfdwtio()+1232<-ksfdwat_internal()+1072<-kfk_reap_ufs_async_io()+224
To view full details, sign in with your My Oracle Support account.
Don't have a My Oracle Support account? Click to get started!
In this Document