On 05/08/2014 08:49 PM, fuyou wrote:
> I'm assuming that's from the driver, I can't tell because the
> stack has been cut off. If so, that's network latency, which is
> time, not cpu-based. Only way to make that better if mysqld is
> configured to do i/o as fast as it can (a topic for
> mysql@stripped <mailto:mysql@stripped>), is to batch
> to cut down on round-trips, so you'll have to get that working.
> Mark Matthews <mark.matthews@stripped
> Consulting Member, Technical Staff - MySQL Enterprise Tools @ Oracle
> MySQL Java Mailing List
> For list archives: http://lists.mysql.com/java
> To unsubscribe: http://lists.mysql.com/java
> the related full jstack :
That's all of the threads, but it's not any full stacks from an
"interesting" thread. However, it looks like you have a lot of threads
waiting on reading results from the database. Because we don't have the
full stack of any of those threads that are waiting on reading from a
socket, we do not know if they are waiting on the results of DML, or if
there's SELECTs going on there, anybody would just be guessing. Can you
characterize the workload at all? What do you expect to happen
performance-wise, and what do you observe? What do various related
metrics on your database server look like? Is the I/O subsystem busy?
idle? overloaded? Are you attempting to insert single rows into a single
table from many, many threads? Is each operation done as auto-commit, or
are multiple statements bracketed by transactions?
Also, please make sure to include the list in your replies.