Re: [AJP] proxy status
At 11:18 PM 8/11/2004, Costin Manolache wrote: William A. Rowe, Jr. wrote: It would be great if we would have at least one or 2 people who are committers in both projects. Since we can't propose Mladen in apache, maybe we can convince Graham to join tomcat :-) Can't propose mturk? Why, because he is already active in httpd space? My point was that it is hard to do development across 2 projects, while none of the developers has CVS access in both. Hopefully that can be addressed... If it is not possible to find a solution to the cvs permissions in either apache or tomcat - then maybe it would be good to have it in a separate repository. It may make it easier to also port it to apache2.0, as a separate standalone module. at one time we had forked mod_proxy development to its own repository, since it was nowhere near ready for 2.0 inclusion. [EMAIL PROTECTED] was created for just this purpose. At least from a development traffic point of view, perhaps it's worth resurrecting that mailing list. As far as code is concerned, Graham, for the moment, is willing to take care of those patches. Although he's subscribed to all three lists, I'd ask that they go either to [EMAIL PROTECTED] or [EMAIL PROTECTED] The history of the discussions is just as important as the actual code commits. Bill - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: [AJP] proxy status
At 11:19 AM 8/12/2004, Justin Erenkrantz wrote: --On Thursday, August 12, 2004 12:57 AM -0500 William A. Rowe, Jr. [EMAIL PROTECTED] wrote: Although he's subscribed to all three lists, I'd ask that they go either to [EMAIL PROTECTED] or [EMAIL PROTECTED] The history of the discussions is just as important as the actual code commits. Can we please not use [EMAIL PROTECTED] As long as the code resides in the main httpd repository, development discussion belongs on [EMAIL PROTECTED], IMHO. That seems rational to me. The reason for proposing [EMAIL PROTECTED] is so that tomcat-dev'ers wouldn't have to swallow the full bandwidth of [EMAIL PROTECTED] (converse of the problem where they asked anyone in [EMAIL PROTECTED] to follow [EMAIL PROTECTED] for the duration of that proxy_ajp development). So I ask our tomcat-dev'ers who are interested in proxy_ajp, proxy_balancer and so on - are you already subscribed/following [EMAIL PROTECTED] Or do you feel a -strong- need for a lower-traffic list? If no one complains loudly, we will keep all proxy traffic on [EMAIL PROTECTED] (cc's to tomcat-dev if you feel a point needs feedback from the tomcat connector folks.) Bill - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: [AJP] proxy status
At 02:15 PM 8/7/2004, Costin wrote: Now let's see how to get this in Apache2.0... Gonna try to make that happen, if I can somehow merge history (ick) It would be great if we would have at least one or 2 people who are committers in both projects. Since we can't propose Mladen in apache, maybe we can convince Graham to join tomcat :-) Can't propose mturk? Why, because he is already active in httpd space? I actually am thinking that generation 2 of mod_aspdotnet will support the ajp connector for out-of-process ASP.NET content ;-) I would like to see ajp discussion back on httpd, or failing that, should we resurrect [EMAIL PROTECTED] for lower bandwidth discussion by both groups? Bill - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: [AJP] proxy status
William A. Rowe, Jr. wrote: At 02:15 PM 8/7/2004, Costin wrote: Now let's see how to get this in Apache2.0... Gonna try to make that happen, if I can somehow merge history (ick) It would be great if we would have at least one or 2 people who are committers in both projects. Since we can't propose Mladen in apache, maybe we can convince Graham to join tomcat :-) Can't propose mturk? Why, because he is already active in httpd space? My point was that it is hard to do development across 2 projects, while none of the developers has CVS access in both. If it is not possible to find a solution to the cvs permissions in either apache or tomcat - then maybe it would be good to have it in a separate repository. It may make it easier to also port it to apache2.0, as a separate standalone module. Costin - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: [AJP] proxy status
Excelent ! Now let's see how to get this in Apache2.0... It would be great if we would have at least one or 2 people who are committers in both projects. Since we can't propose Mladen in apache, maybe we can convince Graham to join tomcat :-) Costin Mladen Turk wrote: Standard: Apache2.0.50/proxy_http 276 req/sec New implementain (DEBUG compile): Apache2.0.50/proxy_http 329 req/sec Apache2.0.50/proxy_ajp 750 req/sec Apache2.0.50/mod_jk 730 req/sec So, we are slihtly faster :) That's probably the double mapping of requests in mod_jk. Costin - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
[AJP] proxy status
Hi all, I've finished the connection pool for mod_proxy. The proxy_http now uses the connection pool (cca 10% speedup for HTTP1.0, the 1.1 should be even higher, but the ab doesn't support 1.1). If someone is willing to test and chase the bugs, he's more then welcome :). TODO: 1. Porting the mechanism from http to proxy_ajp. 2. Using connection pool in ftp. 3. Actually build proxy_balancer [mturk] BUGS: - Still OS 10048)Only one usage of each socket address on WIN32. I'll dig into it to find the reason. - Assertion failed: rl-ntotal == 0, file .\misc\apr_reslist.c, line 164 This manifests on server shutdown. I've probably done some mess with pool cleanup. It would be great if someone finds the cause :) Regards, MT. smime.p7s Description: S/MIME cryptographic signature
Re: [AJP] proxy status
Mladen Turk wrote: Hi all, I've finished the connection pool for mod_proxy. The proxy_http now uses the connection pool (cca 10% speedup for HTTP1.0, the 1.1 should be even higher, but the ab doesn't support 1.1). You should probably test with -k = HTTP/1.0 with keepalive, and request something which sets the content-length. Did you solve the DNS querying issue ? Rémy - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
RE: [AJP] proxy status
Remy Maucherat wrote: I've finished the connection pool for mod_proxy. The proxy_http now uses the connection pool (cca 10% speedup for HTTP1.0, the 1.1 should be even higher, but the ab doesn't support 1.1). You should probably test with -k = HTTP/1.0 with keepalive, and request something which sets the content-length. No the request is still 1.0 and proxy_http checks r-proto_num HTTP_VERSION(1,1). Requested standard servlet/HelloWorldExample. Did you solve the DNS querying issue ? Yes, only the single DNS per worker(connection pool) query is used (unless entire worker is recycled). MT. smime.p7s Description: S/MIME cryptographic signature
Re: [AJP] proxy status
Mladen Turk wrote: I've finished the connection pool for mod_proxy. The proxy_http now uses the connection pool (cca 10% speedup for HTTP1.0, the 1.1 should be even higher, but the ab doesn't support 1.1). If someone is willing to test and chase the bugs, he's more then welcome :). Cool bananas! Possible stupid question: how do I get hold of this code? (Which repository is it in?) TODO: 1. Porting the mechanism from http to proxy_ajp. 2. Using connection pool in ftp. 3. Actually build proxy_balancer [mturk] So to clarify, the proxy framework (ie mod_proxy) now has a connection pool, and mod_proxy_http now uses it (with proxy_ajp on the way)? I am trying to work out which bits are built into proxy, and which bits are in a separate module via a hook. I'd like to see all this stuff go into httpd v2.0 if possible, and want to figure out what the impact will be. There is currently lots of work on mod_cache* and mod_ldap* in httpd, with talk of releasing httpd v2.2 GA around Apachecon time. Regards, Graham -- smime.p7s Description: S/MIME Cryptographic Signature
RE: [AJP] proxy status
Graham Leggett wrote: Cool bananas! Possible stupid question: how do I get hold of this code? (Which repository is it in?) http://cvs.apache.org/viewcvs.cgi/jakarta-tomcat-connectors/ajp/proxy/ Works both on 2.1 and 2.0. TODO: 1. Porting the mechanism from http to proxy_ajp. 2. Using connection pool in ftp. 3. Actually build proxy_balancer [mturk] So to clarify, the proxy framework (ie mod_proxy) now has a connection pool, and mod_proxy_http now uses it (with proxy_ajp on the way)? Yes. I am trying to work out which bits are built into proxy, and which bits are in a separate module via a hook. I'd like to see all this stuff go into httpd v2.0 if possible, and want to figure out what the impact will be. Regarding latest discussinon on HTTP-dev reharding backports, I wish you all the luck :). There is currently lots of work on mod_cache* and mod_ldap* in httpd, with talk of releasing httpd v2.2 GA around Apachecon time. MT. smime.p7s Description: S/MIME cryptographic signature
Re: [AJP] proxy status
Mladen Turk wrote: No the request is still 1.0 and proxy_http checks r-proto_num HTTP_VERSION(1,1). Requested standard servlet/HelloWorldExample.¨ Ok. You can hack a little ab.c to have it generate a HTTP/1.1 request ;) Obviously HTTP/1.0 is not favorable for proxying, so I think HTTP/1.1 would look better. Did you solve the DNS querying issue ? Yes, only the single DNS per worker(connection pool) query is used (unless entire worker is recycled). So the DNS lookup happens once, right ? Good. Rémy - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
RE: [AJP] proxy status
Remy Maucherat wrote: Yes, only the single DNS per worker(connection pool) query is used (unless entire worker is recycled). So the DNS lookup happens once, right ? Good. And... Standard: Apache2.0.50/proxy_http 276 req/sec New implementain (DEBUG compile): Apache2.0.50/proxy_http 329 req/sec MT. smime.p7s Description: S/MIME cryptographic signature
RE: [AJP] proxy status
Standard: Apache2.0.50/proxy_http 276 req/sec New implementain (DEBUG compile): Apache2.0.50/proxy_http 329 req/sec Apache2.0.50/proxy_ajp 750 req/sec Apache2.0.50/mod_jk 730 req/sec So, we are slihtly faster :) MT. smime.p7s Description: S/MIME cryptographic signature
