On Mar 29, 2000, Ekkehard Kraemer wrote:
> Hallo John,
>
> JR>I'm thinking that the kernel TCP connection queues are filling up and
> JR>further requsts are getting dropped, but at 20?
>
> Did you check ulimit (file handles)?
>
> Did you try 'lsof | grep TCP'? Maybe it shows thousands of "zombie" TCP/IP
> connections.
The first thing that I thought of was open file descriptors. lsof
doesn't show any zombies. I'm assuming that all the threads in a
process share the same file descriptors, in which case my server is
only using about 100 file descriptors.
Just in case, I built a new kernel this morning with NR_TASKS,
NR_OPEN, OPEN_MAX and __FD_SETSIZE all bumped up. It performs the
same (actually a little worse).
> Which JDBC driver are you using? The msql driver is (AFAIK)
> unbelievable slow since it creates (and immediately destroys) at
> lease one new thread for each *row* read. By your description,
> JDBC seems not to be the problem, but you may want to check
> anyway.
I've tested this with our own Oracle8i native (OCI) drivers and the
Sybase jConnect type 4 drivers for both Adaptive Server Anywhere and
Adaptive Server Enterprise. I've tried both local database servers
and network based servers (on big hardware).
Is there any kernel output I can enable so I can find out why it
thinks it needs to reject incoming connections?
Thanks
-John
----------------------------------------------------------------------
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
