LCK_ENQUEUE02

[former_member275057]

Hello,

What is the significance of procedure LCK_ENQUEUE02? Is it a new feature of liveCache 7.6.0 and how it would impact on the parallel processing of the jobs?

In case of parallel processing, if we see a lot many tasks with reason "VenterRWregion" and the trace shows that the almost 60% of the runtime is sepnt in LCK_ENQUEUE02, then is it a problem? what could be the next steps of analysis?

Regards,

Satyajit

[former_member229109]

Hello Satyajit,

-> First two questions were discussed with LCA developer & I would like to give you the update with answers.

The LCA routine LCK_ENQUEUE02 functionality is to enqueue the locks.

And you would like to run the "Parallel enqueues". Please pay attention, that the locks server offers two types of granularity for enqueueing locks: each lock is made of an AREA (which is the global scope of the lock, for example, in Demand Planning AREA is the planning area) and of a set of characteristics (a multi-dimensional interval).

The basic idea of using an AREA is to separate lock queries (enqueues, dequeues, ...) into different buckets so that they won't interfere with each other. The lock server works fine in parallel, when most of the lock queries are routed to different AREAS. Within each area, we have to obtain an exclusive lock for the whole AREA, so many parallel queries will block one another and the performance of the lock server goes down.

To get clear on the reported performance problem we would need some additional

information:

(1) In which context the lock server is used? Demand Planning? BI?

(2) How many different AREAS do you have? If you have not enough different AREAS, and most of your queries collide within the same AREA, => it could explain the long runtime.

-> You could overview of the Task States in the database at::

http://dev.mysql.com/doc/maxdb/en/36/d13d408ae01f24e10000000a1550b0/content.htm

liveCache task is in state - VenterRWRegion ::

Task is waiting to enter a protected area (reader/writer region)

The task was trying to get access to a RWRegion, not to a standard region.

-> If you have performance problems in liveCache, and you see many tasks hanging in state "VenterRWregion", could you please run the following commands on the liveCache server to collect additional information:

/sapdb/programs/bin/x_cons <databasename> show all 10 10 > xc_all.prt

/sapdb/programs/bin/x_cons <databasename> show rwreg > xc_rwreg.prt

***

I noticed that you already opened the OSS message 2762565/2007, where further steps for analysis of this issue could be discussed in details.

Thank you and best regards, Natalia Khlopina