CI Locks Causing System Hangs (Doc ID 120654.1)

Problem Description ------------------- You see connections start to hang and CI locks that are hanging the system. You suspect that there is a locking problem. When you check v$lock, you find sessions are waiting for a CI lock. To find out the cause of the problem, you need: i) id1 and id2 from v$lock ii) select sid, event from v$session_wait where wait_time = 0 CI locks are used to invoke specific actions in background processes on a specific instance or all instances. Examples include checkpoint, log switch, shutting down, identifying or re- identifying datafiles, etc. All in all, there are a small number (10's) of predefined cross-instance call types. Oracle 8 Id1 Meaning ~~~ ~~~~~~~ 0 Checkpoint block range 1 Reuse (checkpoint and invalidate) block range 2 LGWR Checkpointing and Hot Backup 3 DBWR syncronization of SGA with control file 4 Log file add/drop/rename notification 5 Write buffer for CR read 6 Test call 7 Invalidate KCK cache in all instances 8 Alter rollback segment optimal 9 Signal Query Servers/coordinator 10 Create Remote parallel query Server 11 Set Global Partitions 12 Stop Disk Writes 13 Drop Sort Segments 14 Release unused space from Sort Segments 15 Instance Recovery for Parallel operation Group 16 Validate parallel slave Lock Value 17 Check transaction state objects 18 Flush blocks in object 19 Rolling release checks 20 Propagate begin backup scn for a file 21 Clear checkpoint progress record id2 Meaning ~~~ ~~~~~~~ 1 Pass in Parameters 2 Invoke the call in background process 3 Foreground has not returned yet 4 Used to allocate the CI call 5 Used to queue up interested clients [For Oracle7] Your CI lock's id1 = 0, and id2 = 5 You are not using Parallel Server and the problem occurs intermittently. Problem Explanation: ==================== One possibility is your applications or packages use dbms_pipe extensively or your sorts are occurring on disk too often. This lock could be used by any foreground and background processes. id1 = 0 means Flush buffers for reuse as new class id2 = 5 means Used to queue up interested clients Solution Description -------------------- 1)Increase the Sort_area_size and make sure your not doing sorts to disk. Sorts to disk cause the CI locks to hang connections. 2)Additionally, pipes can cause this problem The pipes are named memory areas in Oracle's shared pool, where messages are written to and read from. Increasing shared_pool_size and eliminating unused pipes may improve performance. References ---------- Note:1020008.6 SCRIPT: FULLY DECODED LOCKING SCRIPT Additional Search Words ------------------

Related

Products

• Oracle Database Products > Oracle Database Suite > Oracle Database > Oracle Database - Enterprise Edition
• Oracle Database Products > Oracle Database Suite > Oracle Database > Oracle Database - Enterprise Edition > RDBMS

Keywords

TROUBLESHOOT