Re: [JBoss-user] JBoss only using 1 of 8 cpu's
Hi, Something I've noticed is that JBoss actually generates a lot of disk io even though all messages are sent as NON_PERSISTENT. This seems to be mostly in the db/jbossmq/file directory and of temporary nature. -- Alwyn Schoeman SMART Money Inc. "The clock on the wall keeps moving, time stands still... No matter how the dice may fall, someone else always gets to call the number..." msg22208/pgp0.pgp Description: PGP signature
Re: [JBoss-user] JBoss only using 1 of 8 cpu's
Hi, What I'm actually trying to achieve is a log simplere. I am trying simulate a central transaction routing engine using JMS. I know it is a stupid idea, but I need to compare that with using JMS as a system input and RMI for internal computation to make a point. I have 2 setups and I'm not good with ASCII art or any art for that matter so I'll try and explain the design. Setup 1: External PTP client puts JMS ByteMessages onto PTP queue testQueue. MDB takes messages off testQueue and based on state in message routes it to 1 of 4 queues (A,B,C and D). Queues A, B and C has MDB's listening on them and simple puts the message back onto testQueue with a new state. Eventually all messages end up in queue D. Setup 2: External PTP client puts JMS ByteMessages onto PTP queue testQueue. From there it is testQueue->MDB->A->MDB->A2->MDB->B->MDB->B2->MDB->C->MDB->C2->MDB->D. The reason for Setup 2's existence is that 1 uses testQueue for all incoming messages of the routing engine. This means that the load will not be shared effectively between all MDB's. Now, all the MDB's above the put data onto PTP queues have connections open to that queue. This maps to a lot of connections. Is it possible that the problem can be synchronization on these connections at the queue? -- Alwyn Schoeman SMART Money Inc. "The clock on the wall keeps moving, time stands still... No matter how the dice may fall, someone else always gets to call the number..." msg22207/pgp0.pgp Description: PGP signature
Re: [JBoss-user] JBoss only using 1 of 8 cpu's
I saw similar behavior on my 2 cpu Ultra 2 running Linux (Blackdown JVM 1.4.1 beta). It wasn't exclusively using 1 CPU but it almost always had just one java process constantly at 99% cpu utilization. I could perform certain operations that would cause another java process to be runnable. From what I could tell it was the entity bean locking and CMR updates that caused this to happen. The operations I would perform to get another process running were using reporting functions that bypass entity beans (no way I'm going to try to use entity beans to report on 10,000 out of 3million entity beans). My application has been painfully slow because of this so yesterday I completely stripped CMR from the system. This was very easy to do as I have ints as primary keys for all my objects and ejbCreate takes FooData and the CMR objects as parameters... - Change setBar and getBar to set/return int - Add getBarOb() to return the related object - Remove CMR objects from ejbCreate and ejbPostCreate - Fix uses of create to remove CMR objects and call appropriate FooData.set* methods instead - Fix finder methods to take int instead of the object The only tricky part was when I had a finder query that utilitized the CMR relationship to do cross table queries. In my instance I simply implemented the select in JDBC as a home method and returned the ints. Also, making sure to throw ObjectNotFoundException so I didn't have to change logic throughout the code. This works well for simple data types (integer, real, string, etc) which are the same across DBs but not so well for things like boolean (I forcefully map String types to VARCHAR when I need to query them for reporting so that isn't an issue). I had a case where I had a CMR boolean field to check and I chose to leave the finder method intact and perform a check on the objects via getBarOb().getSomeFlag() on the results returned. The short of this is I now have a load of 1.0 instead of 8 on my little box and application response is wonderful. The JVM is also less than 75M where it was around 280M (I set the heap at 256M). And both processors are being used (totaly java process load exceeds 1 cpu when needed). BTW has anyone else noticed Sun JVM 1.4.1 corrupting the heap and crashing on x86 Linux? Brian Macy Fredrik Lindgren wrote: This could also be due to synchronization somewhere in your code. Something to look for is heavy usage of System.out. Writing using System.out is synchronized and can be really slow if you have the console open. I've seen this once on a really powerful AIX cluster (Jboss was not part of that setup though) When tracking the problem we added more trace messages to System.out. That did not help :-) I hope this helped /Fredrik Lindgren Alwyn Schoeman wrote: I get the idea that this is Sparc specific problem, anyone with sparc success? On Fri, Oct 11, 2002 at 09:25:24AM +0200, [EMAIL PROTECTED] wrote: Jboss (3.0.2 at least) is using both of our 2 CPUs on 2.2.* SuSE kernel on intel. On Fri, Oct 11, 2002 at 12:06:23PM +0800, Alwyn Schoeman wrote: Hi, I'm running JBoss 3.0.3 on a Sun Fire 880 with 8 cpu's. Even though I have about 450 threads going it uses only 1 cpu and not very well at that. I have tried all the JVM options for 1.4.1, but no difference. Anyone have any ideas? -- Alwyn Schoeman SMART Money Inc. "The clock on the wall keeps moving, time stands still... No matter how the dice may fall, someone else always gets to call the number..." -- MVH Marius Kotsbak Boost communications AS --- This sf.net email is sponsored by:ThinkGeek Welcome to geek heaven. http://thinkgeek.com/sf ___ JBoss-user mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/jboss-user --- This sf.net email is sponsored by:ThinkGeek Welcome to geek heaven. http://thinkgeek.com/sf ___ JBoss-user mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/jboss-user --- This sf.net email is sponsored by:ThinkGeek Welcome to geek heaven. http://thinkgeek.com/sf ___ JBoss-user mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/jboss-user
RE: [JBoss-user] JBoss only using 1 of 8 cpu's
we've had our 4 cpu sparc box use all CPU's when doing MDB stuff. m -Original Message- From: Alwyn Schoeman [mailto:alwyn@;smart.com.ph] Sent: Thursday, October 10, 2002 9:06 PM To: [EMAIL PROTECTED] Subject: [JBoss-user] JBoss only using 1 of 8 cpu's Hi, I'm running JBoss 3.0.3 on a Sun Fire 880 with 8 cpu's. Even though I have about 450 threads going it uses only 1 cpu and not very well at that. I have tried all the JVM options for 1.4.1, but no difference. Anyone have any ideas? -- Alwyn Schoeman SMART Money Inc. "The clock on the wall keeps moving, time stands still... No matter how the dice may fall, someone else always gets to call the number..." --- This sf.net email is sponsored by:ThinkGeek Welcome to geek heaven. http://thinkgeek.com/sf ___ JBoss-user mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/jboss-user
Re: [JBoss-user] JBoss only using 1 of 8 cpu's
This could also be due to synchronization somewhere in your code. Something to look for is heavy usage of System.out. Writing using System.out is synchronized and can be really slow if you have the console open. I've seen this once on a really powerful AIX cluster (Jboss was not part of that setup though) When tracking the problem we added more trace messages to System.out. That did not help :-) I hope this helped /Fredrik Lindgren Alwyn Schoeman wrote: I get the idea that this is Sparc specific problem, anyone with sparc success? On Fri, Oct 11, 2002 at 09:25:24AM +0200, [EMAIL PROTECTED] wrote: Jboss (3.0.2 at least) is using both of our 2 CPUs on 2.2.* SuSE kernel on intel. On Fri, Oct 11, 2002 at 12:06:23PM +0800, Alwyn Schoeman wrote: Hi, I'm running JBoss 3.0.3 on a Sun Fire 880 with 8 cpu's. Even though I have about 450 threads going it uses only 1 cpu and not very well at that. I have tried all the JVM options for 1.4.1, but no difference. Anyone have any ideas? -- Alwyn Schoeman SMART Money Inc. "The clock on the wall keeps moving, time stands still... No matter how the dice may fall, someone else always gets to call the number..." -- MVH Marius Kotsbak Boost communications AS --- This sf.net email is sponsored by:ThinkGeek Welcome to geek heaven. http://thinkgeek.com/sf ___ JBoss-user mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/jboss-user --- This sf.net email is sponsored by:ThinkGeek Welcome to geek heaven. http://thinkgeek.com/sf ___ JBoss-user mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/jboss-user
Re: [JBoss-user] JBoss only using 1 of 8 cpu's
I get the idea that this is Sparc specific problem, anyone with sparc success? On Fri, Oct 11, 2002 at 09:25:24AM +0200, [EMAIL PROTECTED] wrote: > Jboss (3.0.2 at least) is using both of our 2 CPUs on 2.2.* SuSE kernel on intel. > > On Fri, Oct 11, 2002 at 12:06:23PM +0800, Alwyn Schoeman wrote: > > Hi, > > > > I'm running JBoss 3.0.3 on a Sun Fire 880 with 8 cpu's. Even though I > > have about 450 threads going it uses only 1 cpu and not very well at > > that. I have tried all the JVM options for 1.4.1, but no difference. > > > > Anyone have any ideas? > > > > -- > > Alwyn Schoeman > > SMART Money Inc. > > > > "The clock on the wall keeps moving, time stands still... > > No matter how the dice may fall, someone else always gets to call the number..." > > > > -- > MVH > Marius Kotsbak > Boost communications AS > > > --- > This sf.net email is sponsored by:ThinkGeek > Welcome to geek heaven. > http://thinkgeek.com/sf > ___ > JBoss-user mailing list > [EMAIL PROTECTED] > https://lists.sourceforge.net/lists/listinfo/jboss-user -- Alwyn Schoeman SMART Money Inc. "The clock on the wall keeps moving, time stands still... No matter how the dice may fall, someone else always gets to call the number..." msg22182/pgp0.pgp Description: PGP signature
Re: [JBoss-user] JBoss only using 1 of 8 cpu's
Jboss (3.0.2 at least) is using both of our 2 CPUs on 2.2.* SuSE kernel on intel. On Fri, Oct 11, 2002 at 12:06:23PM +0800, Alwyn Schoeman wrote: > Hi, > > I'm running JBoss 3.0.3 on a Sun Fire 880 with 8 cpu's. Even though I > have about 450 threads going it uses only 1 cpu and not very well at > that. I have tried all the JVM options for 1.4.1, but no difference. > > Anyone have any ideas? > > -- > Alwyn Schoeman > SMART Money Inc. > > "The clock on the wall keeps moving, time stands still... > No matter how the dice may fall, someone else always gets to call the number..." -- MVH Marius Kotsbak Boost communications AS --- This sf.net email is sponsored by:ThinkGeek Welcome to geek heaven. http://thinkgeek.com/sf ___ JBoss-user mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/jboss-user
Re: [JBoss-user] JBoss only using 1 of 8 cpu's
Alwyn Schoeman wrote: > Hi, > > I'm running JBoss 3.0.3 on a Sun Fire 880 with 8 cpu's. Even though I > have about 450 threads going it uses only 1 cpu and not very well at > that. I have tried all the JVM options for 1.4.1, but no difference. > > Anyone have any ideas? > If you are doing web work with Jetty, try upping the Min and Max thread settings in deploy/jbossweb.sar/META-INF/jboss-service.xml. If the concurrency goes up, it might spill over onto more cpus... Jules --- This sf.net email is sponsored by:ThinkGeek Welcome to geek heaven. http://thinkgeek.com/sf ___ JBoss-user mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/jboss-user