Errors encountered during perfromance testing

Question

While doing some stress testing on our WLS server we kept seeing the
following errors:
1) NullPointerException on NTSocketMuxer, error log follows:
Tue Feb 22 05:08:28 EST 2000:<E> <NTSockMux> failure in processSockets()
loop: GetData: fd=16764 numBytes=23
Tue Feb 22 05:08:28 EST 2000:<E> <NTSockMux>
java.lang.NullPointerException: null native pointer - socket was closed
at weblogic.socket.NTSocketMuxer.initiateIO(Native Method)
at weblogic.socket.NTSocketMuxer.processSockets(NTSocketMuxer.java,
Compiled Code)
at
weblogic.socket.SocketReaderRequest.execute(SocketReaderRequest.java:19)
at weblogic.t3.srvr.ExecuteThread.run(ExecuteThread.java, Compiled
Code)
This error doesn't seem to have any immediate impact, although it can't
be a good thing and it seems to happen even at times when we are not
doing stress testing. Also, it is very frequent that we see the above
messages. We are running WLS 4.51 w/ Service Pack 5. I am wondering if
this is a problem w/ Service Pack 5 since I don't see this on another
machine running WLS 4.51 with Service Pack 4. Both are using JDK 1.2.2.
2) Connection Rejected, error log follows:
Tue Feb 22 07:08:16 EST 2000:<W> <ListenThread> Connection rejected:
Login timed out after 15000 msec. The socket came from
[host=207.17.47.141,port=11929,localport=443] See property
weblogic.login.readTimeoutMillis.
I know how to adjust this for this one. My questions on this are:
If I set my readTimeoutMillisSSL (SSL in this case) from 15000 to 30000
what does this exactly mean. Does this mean that instead of allowing a
max 15 seconds for a connection to be established, now I am allowing 30
seconds? Also, is this only for the initial connection establishment (ie
user login), or does this parameter effect other aspects of the
connection later on? What negative side effects would I encounter if I
set this to 60000 (1 minute)?
Finally, what can I do so that a connection does not take over 15
seconds to establish? Note this is not the norm, just happens more often
during stress testing.
3) Creating & Closing connection & DGCserver, log follows:
Tue Feb 22 07:08:44 EST 2000:<I> <RJVM> Closing connection to
138.8.81.19/138.8.81.19 5634583086356155101
Tue Feb 22 07:08:45 EST 2000:<I> <RJVM> Creating connection to
138.8.81.19/138.8.81.19 5634583086356155101
Tue Feb 22 07:08:45 EST 2000:<I> <RJVM> Closing connection to
138.8.81.19/138.8.81.19 5634583086356155101
Tue Feb 22 07:08:50 EST 2000:<I> <RJVM> Creating connection to
144.14.157.204/144.14.157.204 2864292845294268830
Tue Feb 22 07:08:50 EST 2000:<I> <RJVM> Creating connection to
138.8.81.19/138.8.81.19 5634583086356155101
Tue Feb 22 07:08:50 EST 2000:<I> <RJVM> Closing connection to
138.8.81.19/138.8.81.19 5634583086356155101
Tue Feb 22 07:08:50 EST 2000:<I> <RJVM> Creating connection to
138.8.81.19/138.8.81.19 5634583086356155101
Tue Feb 22 07:08:50 EST 2000:<I> <RJVM> Closing connection to
138.8.81.19/138.8.81.19 5634583086356155101
Tue Feb 22 07:08:51 EST 2000:<I> <DGCserver> tried to renew lease for
lost ref: 902
Tue Feb 22 07:08:52 EST 2000:<I> <RJVM> Heartbeat/PublicKey resend
detected
Tue Feb 22 07:08:54 EST 2000:<I> <DGCserver> tried to renew lease for
lost ref: 904
Tue Feb 22 07:08:58 EST 2000:<I> <RJVM> Creating connection to
138.8.81.19/138.8.81.19 5634583086356155101
Why the constant trying to create connection, close connection?
What is DGCserver?
4) Ignoring message from a previous JVM, log follows:
Tue Feb 22 08:29:50 EST 2000:<E> <RJVM> Ignoring message from a previous
generation: JVMMessage from 8729925219143181234C138.8.222.21 to
-3465797227003769874C192.168.100.61 with CMD_ONE_WAY, prtNum=6, ack=103,
seqNum=1384
Tue Feb 22 08:30:15 EST 2000:<I> <HTTPTunneling> Sending DEAD response
What does this mean?
5) PeerGoneExceptions
What causes these?
Our environment is set up as follows:
WLS Server
WLS 4.51 w/ Service Pack 5
NativeIO = true
ExecuteThreadCount = 40
readTimeoutMillis=5000
readTimeoutMillisSSL=10000
Dell Pentium III 600 w/ 512 MB memory
NT 4.0
JavaSoft 1.2.2
-ms128 -mx350
WLS Client
Java Application
t3s and https (using WLS RMI)
JavaSoft 1.1.7b
typically Pentium 200 MHz or better w/ 64MB or more
Basically our clients connect to our WLS server using RMI. Each client
also has a callback object where the server sends event notification
back to the clients. Most of the communication is back through these
client callback objects. Its similar to a stock trading application in
that 1 client incoming requests will generate 200 outgoing events (if
for example there are 200 users on the system). The above observations
where made while 25 very active users where on the system.
Thanks very much for any and all help,
Edwin Marcial
Continental Power Exchange

Answer 1

Hi Kim,
Thanks for the response, but which problem in particular did you solve, I've
listed a couple here.
Edwin
kim hyun chan wrote:
hi,
I met the problem like you before.
so I reduce executeThreadCount from 50 to 20 , and then I solved my proplem.
"Edwin Marcial" <[email protected]> wrote in message
news:[email protected]...
While doing some stress testing on our WLS server we kept seeing the
following errors:
1) NullPointerException on NTSocketMuxer, error log follows:
Tue Feb 22 05:08:28 EST 2000:<E> <NTSockMux> failure in processSockets()
loop: GetData: fd=16764 numBytes=23
Tue Feb 22 05:08:28 EST 2000:<E> <NTSockMux>
java.lang.NullPointerException: null native pointer - socket was closed
at weblogic.socket.NTSocketMuxer.initiateIO(Native Method)
at weblogic.socket.NTSocketMuxer.processSockets(NTSocketMuxer.java,
Compiled Code)
at
weblogic.socket.SocketReaderRequest.execute(SocketReaderRequest.java:19)
at weblogic.t3.srvr.ExecuteThread.run(ExecuteThread.java, Compiled
Code)
This error doesn't seem to have any immediate impact, although it can't
be a good thing and it seems to happen even at times when we are not
doing stress testing. Also, it is very frequent that we see the above
messages. We are running WLS 4.51 w/ Service Pack 5. I am wondering if
this is a problem w/ Service Pack 5 since I don't see this on another
machine running WLS 4.51 with Service Pack 4. Both are using JDK 1.2.2.
2) Connection Rejected, error log follows:
Tue Feb 22 07:08:16 EST 2000:<W> <ListenThread> Connection rejected:
Login timed out after 15000 msec. The socket came from
[host=207.17.47.141,port=11929,localport=443] See property
weblogic.login.readTimeoutMillis.
I know how to adjust this for this one. My questions on this are:
If I set my readTimeoutMillisSSL (SSL in this case) from 15000 to 30000
what does this exactly mean. Does this mean that instead of allowing a
max 15 seconds for a connection to be established, now I am allowing 30
seconds? Also, is this only for the initial connection establishment (ie
user login), or does this parameter effect other aspects of the
connection later on? What negative side effects would I encounter if I
set this to 60000 (1 minute)?
Finally, what can I do so that a connection does not take over 15
seconds to establish? Note this is not the norm, just happens more often
during stress testing.
3) Creating & Closing connection & DGCserver, log follows:
Tue Feb 22 07:08:44 EST 2000:<I> <RJVM> Closing connection to
138.8.81.19/138.8.81.19 5634583086356155101
Tue Feb 22 07:08:45 EST 2000:<I> <RJVM> Creating connection to
138.8.81.19/138.8.81.19 5634583086356155101
Tue Feb 22 07:08:45 EST 2000:<I> <RJVM> Closing connection to
138.8.81.19/138.8.81.19 5634583086356155101
Tue Feb 22 07:08:50 EST 2000:<I> <RJVM> Creating connection to
144.14.157.204/144.14.157.204 2864292845294268830
Tue Feb 22 07:08:50 EST 2000:<I> <RJVM> Creating connection to
138.8.81.19/138.8.81.19 5634583086356155101
Tue Feb 22 07:08:50 EST 2000:<I> <RJVM> Closing connection to
138.8.81.19/138.8.81.19 5634583086356155101
Tue Feb 22 07:08:50 EST 2000:<I> <RJVM> Creating connection to
138.8.81.19/138.8.81.19 5634583086356155101
Tue Feb 22 07:08:50 EST 2000:<I> <RJVM> Closing connection to
138.8.81.19/138.8.81.19 5634583086356155101
Tue Feb 22 07:08:51 EST 2000:<I> <DGCserver> tried to renew lease for
lost ref: 902
Tue Feb 22 07:08:52 EST 2000:<I> <RJVM> Heartbeat/PublicKey resend
detected
Tue Feb 22 07:08:54 EST 2000:<I> <DGCserver> tried to renew lease for
lost ref: 904
Tue Feb 22 07:08:58 EST 2000:<I> <RJVM> Creating connection to
138.8.81.19/138.8.81.19 5634583086356155101
Why the constant trying to create connection, close connection?
What is DGCserver?
4) Ignoring message from a previous JVM, log follows:
Tue Feb 22 08:29:50 EST 2000:<E> <RJVM> Ignoring message from a previous
generation: JVMMessage from 8729925219143181234C138.8.222.21 to
-3465797227003769874C192.168.100.61 with CMD_ONE_WAY, prtNum=6, ack=103,
seqNum=1384
Tue Feb 22 08:30:15 EST 2000:<I> <HTTPTunneling> Sending DEAD response
What does this mean?
5) PeerGoneExceptions
What causes these?
Our environment is set up as follows:
WLS Server
WLS 4.51 w/ Service Pack 5
NativeIO = true
ExecuteThreadCount = 40
readTimeoutMillis=5000
readTimeoutMillisSSL=10000
Dell Pentium III 600 w/ 512 MB memory
NT 4.0
JavaSoft 1.2.2
-ms128 -mx350
WLS Client
Java Application
t3s and https (using WLS RMI)
JavaSoft 1.1.7b
typically Pentium 200 MHz or better w/ 64MB or more
Basically our clients connect to our WLS server using RMI. Each client
also has a callback object where the server sends event notification
back to the clients. Most of the communication is back through these
client callback objects. Its similar to a stock trading application in
that 1 client incoming requests will generate 200 outgoing events (if
for example there are 200 users on the system). The above observations
where made while 25 very active users where on the system.
Thanks very much for any and all help,
Edwin Marcial
Continental Power Exchange