Id verify it by checking the server logs (or number of open sockets) Im not sure what the jmeter error log would look like if it was the jmeter that ran out of file descriptors (for the sockets)
If you cant check the server logs/dont have access to it , an easy to verify is run the same test (but with only half the number of threads) on two client machines i.e. run jmeter from too different machines at the same time. If its a server issue then you should see the same errors. On Tue, Oct 13, 2009 at 3:02 PM, Nikolay_Miroshnichenko <[email protected]>wrote: > Deepak Shetty wrote: > >> Hi >> It looks like your server has reached its limit . you probably need to >> change ulimit on the server. When your requests fail you can probably >> check >> the number of open sockets on the server (using netstat or equivalent). . >> there are other OS related parameters related to how long the socket >> should >> be kept open even after it is closed which affect OS behavior >> >> regards >> deepak >> >> On Tue, Oct 13, 2009 at 2:45 PM, Nikolay_Miroshnichenko >> <[email protected]>wrote: >> >> Hi, >>> we're testing a web site under 1000 users (threads) load. >>> Currently our servers hold about 500 simultaneous users with great >>> latencies. >>> With more users the requests fail. But I'm not sure where the problem is. >>> Whether it the server declines or JMeter or JVM problem. Looking at the >>> error messages in responses (see below) I'd say that JVM (or rather OS, >>> Debian Lenny 64) meets its limits... >>> >>> ***** >>> >>> Thread Name: NewUsers 1-664 >>> Sample Start: 2009-10-10 22:09:31 MSD >>> Load time: 0 >>> Latency: 0 >>> Size in bytes: 1279 >>> Sample Count: 1 >>> Error Count: 1 >>> Response code: Non HTTP response code: java.net.SocketException >>> Response message: Non HTTP response message: Too many open files >>> >>> ***** >>> And a valid answer looks like this: >>> >>> **** >>> >>> Thread Name: NewUsers 1-778 >>> Sample Start: 2009-10-10 22:08:33 MSD >>> Load time: 58159 >>> Latency: 58130 >>> Size in bytes: 611787 >>> Sample Count: 1 >>> Error Count: 0 >>> Response code: 200 >>> Response message: OK >>> >>> ***** >>> >>> Thanks. >>> >>> Nikolay Miroshnichenko >>> >>> >>> >>> >>> --------------------------------------------------------------------- >>> To unsubscribe, e-mail: [email protected] >>> For additional commands, e-mail: [email protected] >>> >>> >> > So it is a server problem, not a test generator. > Thank you! > > -- > Best Regards > Nikolay Miroshnichenko > > > > > --------------------------------------------------------------------- > To unsubscribe, e-mail: [email protected] > For additional commands, e-mail: [email protected] >
