Re: [us...@httpd] Weird timeout connections during benchmark (apache completely stops)?!
asd asd wrote: > That is exactly what you said. Unless you mean keep alives? Yup. > That's > useless I 'm trying to serve requests from different clients, not > multiple requests from the same client. So how is this going to work > with connections getting exhausted at 100? Measuring the 'connect' time is so negligible to be almost immaterial, if you are generating httpd content and attempting to measure the capacity of your server. The exhaustion is not at 100, you merely perceive it to be. You want to combine concurrency (-c) and keepalives (-k) to replicate the processing load on httpd without exhausting the tcp connection id's. - The official User-To-User support forum of the Apache HTTP Server Project. See http://httpd.apache.org/userslist.html> for more info. To unsubscribe, e-mail: users-unsubscr...@httpd.apache.org " from the digest: users-digest-unsubscr...@httpd.apache.org For additional commands, e-mail: users-h...@httpd.apache.org
Re: [us...@httpd] Weird timeout connections during benchmark (apache completely stops)?!
Guys, if i remember currently the default number of concurrent connections served by apache is 256. Right? Shud be good if u tune prefork settings further in httpd-mpm.conf and then rerun your test. Regards Prasanna Ram On Fri, Apr 17, 2009 at 6:07 AM, asd asd wrote: > That is exactly what you said. Unless you mean keep alives? That's useless > I 'm trying to serve requests from different clients, not multiple requests > from the same client. So how is this going to work with connections getting > exhausted at 100? > > > Date: Thu, 16 Apr 2009 18:43:12 -0500 > > From: wr...@rowe-clan.net > > To: users@httpd.apache.org > > Subject: Re: [us...@httpd] Weird timeout connections during benchmark > (apache completely stops)?! > > > > ya...@hotmail.com wrote: > > > So unless I use multiple boxes, I'm not going to be able to service > more > > > than 50-100 requests per second no matter how fast apache is? > > > > Read it again, that's not what I said. READ. Learn. > > > > > > > > - > > The official User-To-User support forum of the Apache HTTP Server > Project. > > See http://httpd.apache.org/userslist.html> for more info. > > To unsubscribe, e-mail: users-unsubscr...@httpd.apache.org > > " from the digest: users-digest-unsubscr...@httpd.apache.org > > For additional commands, e-mail: users-h...@httpd.apache.org > > >
RE: [us...@httpd] Weird timeout connections during benchmark (apache completely stops)?!
That is exactly what you said. Unless you mean keep alives? That's useless I 'm trying to serve requests from different clients, not multiple requests from the same client. So how is this going to work with connections getting exhausted at 100? > Date: Thu, 16 Apr 2009 18:43:12 -0500 > From: wr...@rowe-clan.net > To: users@httpd.apache.org > Subject: Re: [us...@httpd] Weird timeout connections during benchmark (apache > completely stops)?! > > ya...@hotmail.com wrote: > > So unless I use multiple boxes, I'm not going to be able to service more > > than 50-100 requests per second no matter how fast apache is? > > Read it again, that's not what I said. READ. Learn. > > > > - > The official User-To-User support forum of the Apache HTTP Server Project. > See http://httpd.apache.org/userslist.html> for more info. > To unsubscribe, e-mail: users-unsubscr...@httpd.apache.org >" from the digest: users-digest-unsubscr...@httpd.apache.org > For additional commands, e-mail: users-h...@httpd.apache.org >
Re: [us...@httpd] Weird timeout connections during benchmark (apache completely stops)?!
ya...@hotmail.com wrote: > So unless I use multiple boxes, I'm not going to be able to service more > than 50-100 requests per second no matter how fast apache is? Read it again, that's not what I said. READ. Learn. - The official User-To-User support forum of the Apache HTTP Server Project. See http://httpd.apache.org/userslist.html> for more info. To unsubscribe, e-mail: users-unsubscr...@httpd.apache.org " from the digest: users-digest-unsubscr...@httpd.apache.org For additional commands, e-mail: users-h...@httpd.apache.org
Re: [us...@httpd] Weird timeout connections during benchmark (apache completely stops)?!
So unless I use multiple boxes, I'm not going to be able to service more than 50-100 requests per second no matter how fast apache is? -- From: "William A. Rowe, Jr." Sent: Thursday, April 16, 2009 6:27 PM To: Subject: Re: [us...@httpd] Weird timeout connections during benchmark (apache completely stops)?! asd asd wrote: I find that if when concurrent connections are set to anything above 50-100, the server will run fine for 5-10 seconds but then completely stops accepting new connections after that. Not the server, this is TCP/IP 101. Client and server exhaust the number of possible tuples that identify the connection until WAIT_TIMEOUT expires, no more connections are possible. Use keepalives during your stress testing and you'll be much happier with the possible results. - The official User-To-User support forum of the Apache HTTP Server Project. See http://httpd.apache.org/userslist.html> for more info. To unsubscribe, e-mail: users-unsubscr...@httpd.apache.org " from the digest: users-digest-unsubscr...@httpd.apache.org For additional commands, e-mail: users-h...@httpd.apache.org - The official User-To-User support forum of the Apache HTTP Server Project. See http://httpd.apache.org/userslist.html> for more info. To unsubscribe, e-mail: users-unsubscr...@httpd.apache.org " from the digest: users-digest-unsubscr...@httpd.apache.org For additional commands, e-mail: users-h...@httpd.apache.org
Re: [us...@httpd] Weird timeout connections during benchmark (apache completely stops)?!
asd asd wrote: > > I find that if when concurrent connections are set to anything above > 50-100, the server will run fine for 5-10 seconds but then completely > stops accepting new connections after that. Not the server, this is TCP/IP 101. Client and server exhaust the number of possible tuples that identify the connection until WAIT_TIMEOUT expires, no more connections are possible. Use keepalives during your stress testing and you'll be much happier with the possible results. - The official User-To-User support forum of the Apache HTTP Server Project. See http://httpd.apache.org/userslist.html> for more info. To unsubscribe, e-mail: users-unsubscr...@httpd.apache.org " from the digest: users-digest-unsubscr...@httpd.apache.org For additional commands, e-mail: users-h...@httpd.apache.org
Re: [us...@httpd] Weird timeout connections during benchmark (apache completely stops)?!
[Test ~]# httpd -l Compiled in modules: core.c prefork.c http_core.c mod_so.c [Test ~]# httpd -v Server version: Apache/2.2.11 (Unix) Server built: Mar 6 2009 09:12:52 It is a fresh install on fedora 10, with the default configuration benchmarking the default welcome page. Can anyone run some ab/siege and check the results? From: Prasanna Ram Venkatachalam Sent: Thursday, April 16, 2009 11:52 AM To: users@httpd.apache.org Subject: Re: [us...@httpd] Weird timeout connections during benchmark (apache completely stops)?! Could you give more detail about your apache? Also wud be better if u give the details of MPM mode u are using and their configuration details. I dont think there is any limit reg requests specific from external host. Prasanna Ram On Thu, Apr 16, 2009 at 7:47 PM, wrote: apr_socket_recv: Connection reset by peer (104)
Re: [us...@httpd] Weird timeout connections during benchmark (apache completely stops)?!
Could you give more detail about your apache? Also wud be better if u give the details of MPM mode u are using and their configuration details. I dont think there is any limit reg requests specific from external host. Prasanna Ram On Thu, Apr 16, 2009 at 7:47 PM, wrote: > apr_socket_recv: Connection reset by peer (104)
Re: [us...@httpd] Weird timeout connections during benchmark (apache completely stops)?!
The error log is empty! Is there some sort of limit put on number of requests allowed from external host? I just tried ab -n 10 -c 300 http://localhost/ and it did not show the "drop dead" behavior after a certain period of time. However local benchmarking seems to have a ceiling of its own, once I increase the concurrent users to 500 it gives apr_socket_recv: Connection reset by peer (104) But at least 500 is a lot higher than 50. I hear apache is capable of serving thousands of requests per second, what am I missing? Thanks From: Prasanna Ram Venkatachalam Sent: Thursday, April 16, 2009 9:45 AM To: users@httpd.apache.org Subject: Re: [us...@httpd] Weird timeout connections during benchmark (apache completely stops)?! Did the error log say something? It surely will have something incase its not able to find any resources or breached any limits! Regards Prasanna Ram On Thu, Apr 16, 2009 at 5:57 PM, asd asd wrote: I'm trying to get an accurate request/per second of my apache for different pages and for different server options. I am using both ab and siege against apache to try to find out how many users can be loading a specific page at a time. I find that if when concurrent connections are set to anything above 50-100, the server will run fine for 5-10 seconds but then completely stops accepting new connections after that. All connections from ab and siege returns "connection time out". The higher I set the concurrent users, the faster this happens. If I set it to 1000, it happens almost immediately. It is as if the server has ran out on something. But I am not sure what, ran out of sockets? Cpu usageis showing as 2% and plenty of memory. Also this is done over LAN so the network is not the limiting factor. Surely apache must be able to handle over 100 simultaneous connections at the same time? My application will probably have more than 100 users making request at the same time. Any ideas? What command should I be using with ab/siege to properly probe the maximum request/second? PS. so far I am only benching the default page (ie. welcome to apache...)
Re: [us...@httpd] Weird timeout connections during benchmark (apache completely stops)?!
Did the error log say something? It surely will have something incase its not able to find any resources or breached any limits! Regards Prasanna Ram On Thu, Apr 16, 2009 at 5:57 PM, asd asd wrote: > > I'm trying to get an accurate request/per second of my apache for different > pages and for different server options. I am using both ab and siege against > apache to try to find out how many users can be loading a specific page at a > time. > > > I find that if when concurrent connections are set to anything above > 50-100, the server will run fine for 5-10 seconds but then completely stops > accepting new connections after that. All connections from ab and siege > returns "connection time out". The higher I set the concurrent users, the > faster this happens. If I set it to 1000, it happens almost immediately. It > is as if the server has ran out on something. But I am not sure what, ran > out of sockets? Cpu usageis showing as 2% and plenty of memory. Also this is > done over LAN so the network is not the limiting factor. > > > Surely apache must be able to handle over 100 simultaneous connections at > the same time? My application will probably have more than 100 users making > request at the same time. > Any ideas? What command should I be using with ab/siege to properly probe > the maximum request/second? > > > PS. so far I am only benching the default page (ie. welcome to apache...) >
[us...@httpd] Weird timeout connections during benchmark (apache completely stops)?!
I'm trying to get an accurate request/per second of my apache for different pages and for different server options. I am using both ab and siege against apache to try to find out how many users can be loading a specific page at a time. I find that if when concurrent connections are set to anything above 50-100, the server will run fine for 5-10 seconds but then completely stops accepting new connections after that. All connections from ab and siege returns "connection time out". The higher I set the concurrent users, the faster this happens. If I set it to 1000, it happens almost immediately. It is as if the server has ran out on something. But I am not sure what, ran out of sockets? Cpu usageis showing as 2% and plenty of memory. Also this is done over LAN so the network is not the limiting factor. Surely apache must be able to handle over 100 simultaneous connections at the same time? My application will probably have more than 100 users making request at the same time. Any ideas? What command should I be using with ab/siege to properly probe the maximum request/second? PS. so far I am only benching the default page (ie. welcome to apache...)