Applies to:
Oracle Database – Enterprise Edition – Version 12.1.0.1 to 12.1.0.2 [Release 12.1]
Oracle Database Cloud Schema Service – Version N/A and later
Gen 1 Exadata Cloud at Customer (Oracle Exadata Database Cloud Machine) – Version N/A and later
Oracle Cloud Infrastructure – Database Service – Version N/A and later
Oracle Database Cloud Exadata Service – Version N/A and later
IBM AIX on POWER Systems (64-bit)

Description
Oracle 12c introduces a new default feature of using multiple LGWRs which may lead to DEADLOCK / Database Hang or ORA-742 “Log read detects lost write” or ORA-600 [kcrfrgv_nextlwn_scn] during instance OPEN or ORA-600 [krr_process_read_error_2] during Recovery on IBM AIX and potentially on HPUX Itanium 64bit.
The database may become unusable and fail to be OPEN.

Occurrence
This issue is specific to RDBMS version 12c (12.1.0.1 or 12.1.0.2) where the new default feature of using multiple LGWRs is introduced.
It affects databases on IBM AIX and potentially on HPUX Itanium 64bit

Symptoms
ORACLE on IBM AIX or HPUX Itanium 64bit with RDBMS Version 12c.

DEADLOCK or ORA-742 “Log read detects lost write” or ORA-600 [kcrfrgv_nextlwn_scn] during instance OPEN or ORA-600 [krr_process_read_error_2] during Recovery caused by bug 21915719.

PMON may terminate the instance while extensive block recovery is being performed.

A DEADLOCK example is with LG0[n] waiting on ‘LGWR worker group ordering’. Example from a System State Dump trace file:

PROCESS 18: LG01
SO: 0x7000101f95ad720, type: 4, owner: 0x7000101f84195f8, flag: INIT/-/-/0x00
if: 0x3 c: 0x3
   proc=0x7000101f84195f8, name=session, file=ksu.h LINE:13590 ID:, pg=0
conuid=0
  (session) sid: 865 ser: 1 trans: 0x0, creator: 0x7000101f84195f
 Current Wait Stack:
   0: waiting for 'LGWR worker group ordering'
      lwn_id=0x58, phase=0x1, =0x0
      wait_id=4947 seq_num=4948 snap_id=1
      wait times: snap=13 min 21 sec, exc=13 min 21 sec, total=13 min 21 sec
      wait times: max=infinite, heur=13 min 21 sec
      wait counts: calls=1 os=267
      in_wait=1 iflags=0x5a0
  There is at least one session blocking this session.
    Dumping 1 direct blocker(s):
      inst: 1, sid: 817, ser: 1
    Dumping final blocker:
      inst: 1, sid: 817, ser: 1
  There are 730 sessions blocked by this session.
.
.
PROCESS 17: LG00
SO: 0x7000101f85bcc60, type: 4, owner: 0x7000101f93eeb20, flag: INIT/-/-/0x00
if: 0x3 c: 0x3
   proc=0x7000101f93eeb20, name=session, file=ksu.h LINE:13590 ID:, pg=0
conuid=0
  (session) sid: 817 ser: 1 trans: 0x0, creator: 0x7000101f93eeb20
  ksuxds FALSE at location: 0
  service name: SYS$BACKGROUND
  Current Wait Stack:
   0: waiting for 'LGWR worker group ordering'
      lwn_id=0x56, phase=0x1, =0x0
      wait_id=1630680 seq_num=57841 snap_id=1
      wait times: snap=13 min 21 sec, exc=13 min 21 sec, total=13 min 21 sec
      wait times: max=infinite, heur=13 min 21 sec
      wait counts: calls=2 os=268
      in_wait=1 iflags=0x15a0
  There is at least one session blocking this session.
    Dumping 1 direct blocker(s):
      inst: 1, sid: 865, ser: 1
    Dumping final blocker:
      inst: 1, sid: 865, ser: 1
 

The instance may fail to OPEN with errors ORA-600 [kcrfrgv_nextlwn_scn] and/or ORA-600 [krr_process_read_error_2]:

Recovery Session Failed with:

ORA-00283: recovery session canceled due to errors
ORA-00600: internal error code, arguments: [krr_process_read_error_2],


Alter database open fails with:

ORA-00600: internal error code, arguments: [kcrfrgv_nextlwn_scn] .....
ORA-600 signalled during: ALTER DATABASE OPEN...

Workaround
Disable the new feature of multiple LGWR worker processes by proactively setting _use_single_log_writer=true.

Setting _use_single_log_writer = true is a safe workaround; it is the behavior before 12c where multiple LGWR worker groups were not available.

ALTER SYSTEM SET "_use_single_log_writer"=TRUE SID='*' SCOPE=SPFILE;
-- Restart the database or all instances of the RAC database

Note that while _use_single_log_writer=true is not set, then error ORA-600 [kcrfrgv_nextlwn_scn] might be produced avoiding the database to OPEN. Once the problem is introduced, _use_single_log_writer=true may not fix it. _use_single_log_writer = true prevents inconsistencies in the redo log to be introduced which causes that error.
If the parameter does not help, because the problem was already introduced when _use_single_log_writer=true had not been proactively set, then Point in Time Recovery (PITR) or Flashback Database are the options to recover from this situation.
参考:ALERT: Bug 21915719 Database hang or may fail to OPEN in 12c IBM AIX or HPUX Itanium – ORA-742, DEADLOCK or ORA-600 [kcrfrgv_nextlwn_scn] ORA-600 [krr_process_read_error_2] (Doc ID 1957710.1)