# Do NOT uncomment this line as it causes in VM calls to go over RMI!
#java.naming.provider.url="">
So it looks like setting it to jnp://localhost:1099 could really be adding to the socket load on the box, and thus we could be running out of file descriptors after a week or so, which is roughly how long we can stay running. Does this sound plausible?
As far as stack size, our default is 8m. Does anyone have any recommendations on what we should set this too? Or suggestions on determining what is appropriate for our particular situation?
thanks.
.peter
Adrian Brock wrote:
On Thu, 2003-12-04 at 21:15, Peter Luttrell wrote:We're using JBoss3.2.1 with Jetty on RedHat 9 with Suns 1.4.2_01 vm. We have a pretty heavy load.After roughly a week many of the boxes start to experience weird problems where JBoss is unable to get what looks to be socket connections. In some cases, we cannot contact our mailserver, in other cases we cannot contact our database; in the latest case we're unable to get a connection to the local jndi server (localhost:1099). Sometimes a simple restart of jboss will sometimes solves the problems, othertimes we have to restart linux. The times of the crashes are roughly 4am and sometimes 6am, so it's likely caused by a cron job running at those times, which we're currently looking into. Has anyone experienced similar problems? .peterThis sounds like a problem with sockets. Beyond what John said about hsqldb, use "netstat -an" to see what sockets are open and what state they are in. But you should also check that the vm is responding at all and hasn't just hung. Use kill -3 <jboss-pid> to see whether it produces a thread dump. The most likely problem is your access to jndi. Why are you making it use a socket at all? It should be using in memory access. Either you have changed conf/jndi.properties (don't) or you have a jndi.properties deployed somewhere else with a provider url specified. jndi access should not be using a local socket inside jboss. Unrelated, you might also want to reduce the stack size. The typical 8M of stack is very large for java which allocates most stuff on the heap. Use ulimit -a to see your default setting. NOTE: There is a ulimit command in bin/run.sh where you should change this. Regards, Adrian---------- CONFIDENTIALITY NOTICE: This e-mail message, including any attachments,is for the sole use of the intended recipient(s), even if addressed incorrectly, and may contain confidential and privileged information. Any unauthorized review, use, disclosure or distribution is prohibited. If you are not the intended recipient, please contact the sender by reply e-mail and destroy or delete all copies of the original message and all attachments, including deletion from the trash or equivalent folder. Thank you.
