[Search for users] [Overall Top Noters] [List of all Conferences] [Download this site]

Conference utrop1::linkworks_v3

Title:	LinkWorks V3.0 Notes Conference
Notice:	LNX_APO = APO issues, LINKWORKS_V3 = V3.0 issues
Moderator:	tacklr.apd.dec.com::TACK_Lm::TACK_L

Created:	Tue Jun 28 1994
Last Modified:	Fri Jun 06 1997
Last Successful Update:	Fri Jun 06 1997
Number of topics:	2269
Total number of notes:	8338

2229.0. "LinkWorks 3.0.7 hangs with Oracle and OSF/1" by BACHUS::dbc624.bro.dec.com::POPPE () Mon May 12 1997 18:02

Hi,

A customer with some 300 users on LinkWorks V3.0.7 reports some hang 
situations from time to time: all LinkWorks users have to wait (over 10 
minutes), while there is almost no CPU activity on the server.  Only Oracle is 
writing to disk (ora_dbwr_SID).

I asked to check database parameters (SGA size, nbr of rollback segments and 
size, frequency of log switches ...).  

The customer told me they use 'register interest' rather intensively:
They have created a folder that is used as a bulletin board, and all users 
have registered interest on this.  A few people are busy updating this 
bulletin board.

How can he find out if this causes the hangs ?
Are there specific database tables that can be monitored for this, e.g. by
checking the row count ?
Is there a way to find out what request LinkWorks has made to the database at
a given moment (i.e. when this hang occurs) ?

B.t.w. I asked to check for a TDF hang, but they never had to stop and restart 
LinkWorks, it allways continued after some time.

Regards,
Marc

T.R Title User Personal
Name Date Lines

2229.1 elec1.uto.dec.com::wijermars_v Tue May 20 1997 11:08 43

T.R	Title	User	Personal Name	Date	Lines
2229.1		elec1.uto.dec.com::wijermars_v		`Tue May 20 1997 11:08`	43
	Hi, I think Oracle is busy during that time,blocking all activity to the database. I might be interesting having a look at Oracle statistics. See the Oracle tuning manual for this. Some interesting queries might be: (I just took them from the tuning manual) Monitor statistics in v$librarycache > select sum(pins) "Executions", sum(reloads) "Cache Misses" from v$librarycache; Miss ratio should be less than 1% If not, increase SHARED_POOL_SIZE in init<SID>.ora Monitor statistics in v$rowcache > select sum(gets) "Gets", sum(getmisses) "Get Misses" from v$rowcache; Miss ratio should be less than 15% If not, increase SHARED_POOL_SIZE in init<SID>.ora Monitor statistics from v$sysstat > select name, value from v$sysstat where name in (�db block gets�, �consistent gets�, �physical reads�); Hitratio=1-(physical reads / (db block gets + consistent gets) Hitratio should be above 60-70% If not, increase DB_BLOCK_BUFFERS in init<SID>.ora Reduce contention on rollback segments Monitor v$waitstat >select class, count from v$waitstat where class in (�system undo header�, �system undo block�, �undo header�, �undo block�); Check total number of request > select sum(value) from v$sysstat where name in (�db block gets�, �consistent gets�); Number of waits for each class should be less than 1% If not, create more rollback segments > create rollback segment rb5 tablespace rbs storage(...); Vincent

Hi,

I think Oracle is busy during that time,blocking all activity to the database.
I might be interesting having a look at Oracle statistics.

See the Oracle tuning manual for this.
Some interesting queries might be: (I just took them from the tuning manual)

Monitor statistics in v$librarycache
> select sum(pins) "Executions", sum(reloads) "Cache Misses"  from v$librarycache;
	
Miss ratio should be less than 1%
If not, increase SHARED_POOL_SIZE in init<SID>.ora

Monitor statistics in v$rowcache
> select sum(gets) "Gets", sum(getmisses) "Get Misses" from v$rowcache;

Miss ratio should be less than 15%
If not, increase SHARED_POOL_SIZE in init<SID>.ora

Monitor statistics from v$sysstat
> select name, value from v$sysstat where name in (�db block gets�, �consistent gets�, �physical 
reads�);

Hitratio=1-(physical reads / (db block gets + consistent gets)
Hitratio should be above 60-70%
If not, increase DB_BLOCK_BUFFERS in init<SID>.ora

Reduce contention on rollback segments
Monitor v$waitstat
>select class, count from v$waitstat where class in (�system undo header�, �system undo block�, 
�undo header�, �undo block�);

Check total number of request
> select sum(value) from v$sysstat where  name in (�db block gets�, �consistent gets�);

Number of waits for each class should be less than 1%
If not, create more rollback segments
> create rollback segment rb5 tablespace rbs storage(...);



Vincent