onstat -s Latches

The onstat -s command displays the latches that are currently being held by userthreads for resources in shared memory.

Latches with lock or userthread set
name     address           lock wait userthread
physlog  4ebdf368         0    3dbc77d20        3dca48d20
 physb1  4ebdf3e0         0    3dca48d20        3dc7bdaa8
 physb2  4ebdf4e0         0    0                3dca48d20

</div>

Output Description

Heading	Description	Format	See Also
name	The name of the resource upon which a latch has been acquired. Name Description altlatch Alter table counter archive Archive bf[#] Buffer bh[#] Hash buffer chunks Chunks ckpt Checkpoints DBSpace DBSpace flushctl Flusher control flushr# Page cleaners locks Lock table lh[#] Lock hash lockfr Lock free list lockdl Deadlock list loglog Logical Log LRU# LRU queue physb1 Physical Log buffer 1 physb2 Physical Log buffer 2 physlog Physical Log pt Tblspace tblspace tblsps Tblspace table timestamp Timestamp trace Trace buffer trans Transactions txlk[#] Transaction list users User table userthreads User list </div>	Str	onstat -u onstat -g amx onstat -g lmx onstat -g wmx
address	The in-memory address of the latch.	Hex
lock	Indicates if the lock on the resource is set. The code that indicates the lock status (1 or 0) is machine dependent	Dec
wait	The first address of the thread waiting for this resource.	Hex
userthread	The address of the userthread holding the latch.	Hex	onstat -u

Notes

Latches are a method for controlling simultaneous access to shared memory resources.

The userthread address can be used with onstat -u command to track the user

To obtain the rstcb address from the tcb address, examine the output of the onstat -g ath command, which lists both addresses for each user thread.

Monitoring and Tuning

Latches are grabbed and released very rapidly. As such, you generally will not see the same latch being held by the same thread if you run the onstat command a second time.

The number of latches (mutexes) required can be reduced by using one CPU VP and setting the value of SINGLE_CPU_VP to 1 in the configuration file when working on a one or two CPU machine.

Another method for monitoring mutexes is through the MT commands amx, lmx and wmx, which show all mutexes, locked mutexes and mutexes with waiters.