This is the mail archive of the ecos-discuss@sources.redhat.com mailing list for the eCos project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]

TCP/IP Stack and OUT of MBUFS ( Clusters )


Hi all,

I am having a problem with the TCP/IP stack. I am doing some "stress
testing"
with the eSoap ( SOAP stack ) for ecos, and after a little while ( a couple
of minutes, sometimes less ), I ran out of MBUFS and Mclusters.

I have the latest "super duper" updates from CVS.

I have modifed the "server_test.c" ( tcp/ip tests ) and hacked a simple
java client that "re-create" the problem. I would appreaciate if anyone
could take a look at it. It won't take more than 2 minutes. :-)

I am not sure if I am "hitting" a bug or it is just a config issue.

I cannot believe that my demo only works for a couple of minutes,
and the target freezes after that( for TCP/IP communication ).

I have attached some of logs I saved, and the sources for the tests.

It looks like we are running out of Mclusters ( whatever that means ).

I would appreaciate any suggestions about what I have to do.
It maybe a config problem, or maybe I am just overloading
the stack with too many packets.

Things get a lot worse when I use a tool for stress testing to
simulate requests from many clients, such as:

http://webtool.rte.microsoft.com/

Thanks for your time.

Rosimildo.




















mbuf_problem.zip


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]