db-derby-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From hofhansl <lhofha...@yahoo.com>
Subject Re: Derby process taking up more than 20% of cpu for more than 5hrs with only one user connected.
Date Sat, 21 Apr 2007 00:29:06 GMT

If another process is constantly updating the table, the reads may not be
successful in getting a read-lock and eventually time out.

You can try 
> select count(*) from M_20070129MESSAGES WITH UR;

WITH UR allows for uncommitted reads.
You may also have to increase the lock escalation threshold (number of locks
after which Derby decides to get a table lock). Something like:
derby.locks.escalationThreshold=10000 in derby.properties.


Sachin Chitale wrote:
> 
> Hi,
> 
> To give you a background of what I was testing:
> 
> We have process that uses database as our guranteed message store. To give
> a 
> stress test to Derby, this program was run to store about 70k+ messages.
> Then 
> a subscriber was started to fetch the messages. But it seems that the
> database 
> has stopped responding to the requests.
> 
> Initially I was able to get message count using:
> ij>  select count(*) from M_20070129MESSAGES;
> 1
> -----------
> 54598
> 
> After ~15 minutes.
> 1 row selected
> ij> select count(*) from M_20070129MESSAGES;
> ERROR 40XL1: A lock could not be obtained within the time requested
> 
> 

-- 
View this message in context: http://www.nabble.com/Derby-process-taking-up-more-than-20--of-cpu-for-more-than-5hrs-with-only-one-user-connected.-tf3619391.html#a10112431
Sent from the Apache Derby Users mailing list archive at Nabble.com.


Mime
View raw message