Re: Does mod_perl/mod_??? need a hook called when a request/conn leaves the original worker thread?
On Sat, Nov 30, 2013 at 10:03 AM, Jeff Trawick wrote: > On Tue, Nov 26, 2013 at 3:17 PM, Jeff Trawick wrote: > >> On Tue, Nov 26, 2013 at 9:01 AM, Jeff Trawick wrote: >> >>> On Tue, Nov 26, 2013 at 8:55 AM, Graham Leggett wrote: >>> On 26 Nov 2013, at 3:51 PM, Jeff Trawick wrote: > As it turns out (or, why didn't I refresh my understanding before), the MPM only knows about the conn_rec. > > * It could do extra work to learn about the request in order to pass the request to the new hook. > * It could avoid that extra work for configurations that don't have a module that implements the hook. > > I'm leaning towards not having the MPM bother with any of that. Such magic is well within the scope of a module that cares about detaching from the thread anyway. It would be nice if there was a clean and consistent way for c->output_filters to become r->output_filters, and when the request is cleaned up for the c->output_filters to be reverted back to what it was before. This way content and resource filters could take advantage of write completion in future. >>> >>> Interesting... I don't know exactly what "clean ... way" will mean in >>> this context, but I can look at early-request-hook+request-pool-cleanup >>> processing to let the MPM track the current r for a connection. >>> >>> Regards, Graham -- >>> >>> >> Here's a first draft for suspend/resume hooks: >> >> http://people.apache.org/~trawick/suspend_resume_hooks_r1.txt >> >> This maintains r inside the event MPM connection state. >> >> From a module that tries to log r->the_request when the connection is >> suspended or resumed: >> >> [pid 31968:139866574595840] suspend, r 7f3530002970 GET >> /ubuntu-12.04.3-desktop-i386.iso HTTP/1.1 >> [pid 31968:139866566203136] resume, r 7f3530002970 GET >> /ubuntu-12.04.3-desktop-i386.iso HTTP/1.1 >> [pid 31968:139866566203136] suspend, r 7f3530002970 GET >> /ubuntu-12.04.3-desktop-i386.iso HTTP/1.1 >> [pid 31968:139866557810432] resume, r 7f3530002970 GET >> /ubuntu-12.04.3-desktop-i386.iso HTTP/1.1 >> [pid 31968:139866557810432] suspend, r 7f3530002970 GET >> /ubuntu-12.04.3-desktop-i386.iso HTTP/1.1 >> [pid 31968:139866549417728] resume, r 7f3530002970 GET >> /ubuntu-12.04.3-desktop-i386.iso HTTP/1.1 >> ... >> [pid 31968:139866725664512] suspend, r 0 >> >> (For detecting the end of the request or connection, the module needs to >> use a cleanup on the appropriate pool as always.) >> >> Todos (that I know of so far :) ): >> >> 1. call the new hooks from places other than process_socket() >> > > AFAICT, the other place where conn-specific* non-MPM code can run is in > pool cleanups, so resume may have to be called in a "pre-cleanup", if the > conn is currently suspended. > > *There are special callbacks handled by event[/eventopt] but they're not > associated with the client connection conn_rec so they are not the target > of the suspend/resume hooks. > > >> 2. consider passing a coarse, non-MPM-specific, representation of the >> state on the hook calls >> > > I didn't realize that the conn_rec already has a very fine-grained > representation of the state. This state info is sufficient and, as it is > already part of the API, there's no concern at this point in creating a > coarse representation that is easier for multiple MPMs to maintain without > breaking the API when the MPM implementation changes. > Implemented for Event in r1546759; I'll implement for eventopt before long; it would be great if Event-savvy folks would take a look ;) -- Born in Roswell... married an alien... http://emptyhammock.com/
Re: Does mod_perl/mod_??? need a hook called when a request/conn leaves the original worker thread?
On Tue, Nov 26, 2013 at 3:17 PM, Jeff Trawick wrote: > On Tue, Nov 26, 2013 at 9:01 AM, Jeff Trawick wrote: > >> On Tue, Nov 26, 2013 at 8:55 AM, Graham Leggett wrote: >> >>> On 26 Nov 2013, at 3:51 PM, Jeff Trawick wrote: >>> >>> > As it turns out (or, why didn't I refresh my understanding before), >>> the MPM only knows about the conn_rec. >>> > >>> > * It could do extra work to learn about the request in order to pass >>> the request to the new hook. >>> > * It could avoid that extra work for configurations that don't have a >>> module that implements the hook. >>> > >>> > I'm leaning towards not having the MPM bother with any of that. Such >>> magic is well within the scope of a module that cares about detaching from >>> the thread anyway. >>> >>> It would be nice if there was a clean and consistent way for >>> c->output_filters to become r->output_filters, and when the request is >>> cleaned up for the c->output_filters to be reverted back to what it was >>> before. >>> >>> This way content and resource filters could take advantage of write >>> completion in future. >>> >> >> Interesting... I don't know exactly what "clean ... way" will mean in >> this context, but I can look at early-request-hook+request-pool-cleanup >> processing to let the MPM track the current r for a connection. >> >> >>> >>> Regards, >>> Graham >>> -- >>> >>> >> >> > Here's a first draft for suspend/resume hooks: > > http://people.apache.org/~trawick/suspend_resume_hooks_r1.txt > > This maintains r inside the event MPM connection state. > > From a module that tries to log r->the_request when the connection is > suspended or resumed: > > [pid 31968:139866574595840] suspend, r 7f3530002970 GET > /ubuntu-12.04.3-desktop-i386.iso HTTP/1.1 > [pid 31968:139866566203136] resume, r 7f3530002970 GET > /ubuntu-12.04.3-desktop-i386.iso HTTP/1.1 > [pid 31968:139866566203136] suspend, r 7f3530002970 GET > /ubuntu-12.04.3-desktop-i386.iso HTTP/1.1 > [pid 31968:139866557810432] resume, r 7f3530002970 GET > /ubuntu-12.04.3-desktop-i386.iso HTTP/1.1 > [pid 31968:139866557810432] suspend, r 7f3530002970 GET > /ubuntu-12.04.3-desktop-i386.iso HTTP/1.1 > [pid 31968:139866549417728] resume, r 7f3530002970 GET > /ubuntu-12.04.3-desktop-i386.iso HTTP/1.1 > ... > [pid 31968:139866725664512] suspend, r 0 > > (For detecting the end of the request or connection, the module needs to > use a cleanup on the appropriate pool as always.) > > Todos (that I know of so far :) ): > > 1. call the new hooks from places other than process_socket() > AFAICT, the other place where conn-specific* non-MPM code can run is in pool cleanups, so resume may have to be called in a "pre-cleanup", if the conn is currently suspended. *There are special callbacks handled by event[/eventopt] but they're not associated with the client connection conn_rec so they are not the target of the suspend/resume hooks. > 2. consider passing a coarse, non-MPM-specific, representation of the > state on the hook calls > I didn't realize that the conn_rec already has a very fine-grained representation of the state. This state info is sufficient and, as it is already part of the API, there's no concern at this point in creating a coarse representation that is easier for multiple MPMs to maintain without breaking the API when the MPM implementation changes. -- Born in Roswell... married an alien... http://emptyhammock.com/
Re: Does mod_perl/mod_??? need a hook called when a request/conn leaves the original worker thread?
On Tue, Nov 26, 2013 at 9:01 AM, Jeff Trawick wrote: > On Tue, Nov 26, 2013 at 8:55 AM, Graham Leggett wrote: > >> On 26 Nov 2013, at 3:51 PM, Jeff Trawick wrote: >> >> > As it turns out (or, why didn't I refresh my understanding before), the >> MPM only knows about the conn_rec. >> > >> > * It could do extra work to learn about the request in order to pass >> the request to the new hook. >> > * It could avoid that extra work for configurations that don't have a >> module that implements the hook. >> > >> > I'm leaning towards not having the MPM bother with any of that. Such >> magic is well within the scope of a module that cares about detaching from >> the thread anyway. >> >> It would be nice if there was a clean and consistent way for >> c->output_filters to become r->output_filters, and when the request is >> cleaned up for the c->output_filters to be reverted back to what it was >> before. >> >> This way content and resource filters could take advantage of write >> completion in future. >> > > Interesting... I don't know exactly what "clean ... way" will mean in > this context, but I can look at early-request-hook+request-pool-cleanup > processing to let the MPM track the current r for a connection. > > >> >> Regards, >> Graham >> -- >> >> > > Here's a first draft for suspend/resume hooks: http://people.apache.org/~trawick/suspend_resume_hooks_r1.txt This maintains r inside the event MPM connection state. >From a module that tries to log r->the_request when the connection is suspended or resumed: [pid 31968:139866574595840] suspend, r 7f3530002970 GET /ubuntu-12.04.3-desktop-i386.iso HTTP/1.1 [pid 31968:139866566203136] resume, r 7f3530002970 GET /ubuntu-12.04.3-desktop-i386.iso HTTP/1.1 [pid 31968:139866566203136] suspend, r 7f3530002970 GET /ubuntu-12.04.3-desktop-i386.iso HTTP/1.1 [pid 31968:139866557810432] resume, r 7f3530002970 GET /ubuntu-12.04.3-desktop-i386.iso HTTP/1.1 [pid 31968:139866557810432] suspend, r 7f3530002970 GET /ubuntu-12.04.3-desktop-i386.iso HTTP/1.1 [pid 31968:139866549417728] resume, r 7f3530002970 GET /ubuntu-12.04.3-desktop-i386.iso HTTP/1.1 ... [pid 31968:139866725664512] suspend, r 0 (For detecting the end of the request or connection, the module needs to use a cleanup on the appropriate pool as always.) Todos (that I know of so far :) ): 1. call the new hooks from places other than process_socket() 2. consider passing a coarse, non-MPM-specific, representation of the state on the hook calls -- Born in Roswell... married an alien... http://emptyhammock.com/
Re: Does mod_perl/mod_??? need a hook called when a request/conn leaves the original worker thread?
On Tue, Nov 26, 2013 at 8:55 AM, Graham Leggett wrote: > On 26 Nov 2013, at 3:51 PM, Jeff Trawick wrote: > > > As it turns out (or, why didn't I refresh my understanding before), the > MPM only knows about the conn_rec. > > > > * It could do extra work to learn about the request in order to pass the > request to the new hook. > > * It could avoid that extra work for configurations that don't have a > module that implements the hook. > > > > I'm leaning towards not having the MPM bother with any of that. Such > magic is well within the scope of a module that cares about detaching from > the thread anyway. > > It would be nice if there was a clean and consistent way for > c->output_filters to become r->output_filters, and when the request is > cleaned up for the c->output_filters to be reverted back to what it was > before. > > This way content and resource filters could take advantage of write > completion in future. > Interesting... I don't know exactly what "clean ... way" will mean in this context, but I can look at early-request-hook+request-pool-cleanup processing to let the MPM track the current r for a connection. > > Regards, > Graham > -- > > -- Born in Roswell... married an alien... http://emptyhammock.com/
Re: Does mod_perl/mod_??? need a hook called when a request/conn leaves the original worker thread?
On 26 Nov 2013, at 3:51 PM, Jeff Trawick wrote: > As it turns out (or, why didn't I refresh my understanding before), the MPM > only knows about the conn_rec. > > * It could do extra work to learn about the request in order to pass the > request to the new hook. > * It could avoid that extra work for configurations that don't have a module > that implements the hook. > > I'm leaning towards not having the MPM bother with any of that. Such magic > is well within the scope of a module that cares about detaching from the > thread anyway. It would be nice if there was a clean and consistent way for c->output_filters to become r->output_filters, and when the request is cleaned up for the c->output_filters to be reverted back to what it was before. This way content and resource filters could take advantage of write completion in future. Regards, Graham --
Re: Does mod_perl/mod_??? need a hook called when a request/conn leaves the original worker thread?
On Mon, Nov 18, 2013 at 12:15 PM, Eric Covener wrote: > On Mon, Nov 18, 2013 at 10:58 AM, Jeff Trawick wrote: > > For the mod_perl crash with Event that I posted at the URL below, I would > > suspect that there's some affinity with the original worker thread. Can > > anyone in mod_perl land confirm? > > > > > http://mail-archives.apache.org/mod_mbox/perl-dev/201311.mbox/%3CCAKUrXK6C3R_F3NdA%2BJUGYOqppvnoQJLTGQ9%2BA916vuMb0g9dig%40mail.gmail.com%3E > > > > I'm also aware of a third-party diagnostic module that could use a hint, > and > > in general I wonder if anyone knows of specific interface requirements > that > > would need to be provided by a new hook for indicating when a connection > or > > request leaves the original thread or is handled by a new one. > > I can't see any need for more than request_rec on both ends. > As it turns out (or, why didn't I refresh my understanding before), the MPM only knows about the conn_rec. * It could do extra work to learn about the request in order to pass the request to the new hook. * It could avoid that extra work for configurations that don't have a module that implements the hook. I'm leaning towards not having the MPM bother with any of that. Such magic is well within the scope of a module that cares about detaching from the thread anyway. -- Born in Roswell... married an alien... http://emptyhammock.com/
Re: Does mod_perl/mod_??? need a hook called when a request/conn leaves the original worker thread?
Ouch, I meant to sent this to dev@perl instead of dev@apr... On Mon, Nov 18, 2013 at 10:58 AM, Jeff Trawick wrote: > For the mod_perl crash with Event that I posted at the URL below, I would > suspect that there's some affinity with the original worker thread. Can > anyone in mod_perl land confirm? > > > http://mail-archives.apache.org/mod_mbox/perl-dev/201311.mbox/%3CCAKUrXK6C3R_F3NdA%2BJUGYOqppvnoQJLTGQ9%2BA916vuMb0g9dig%40mail.gmail.com%3E > > I'm also aware of a third-party diagnostic module that could use a hint, > and in general I wonder if anyone knows of specific interface requirements > that would need to be provided by a new hook for indicating when a > connection or request leaves the original thread or is handled by a new one. > -- > Born in Roswell... married an alien... > http://emptyhammock.com/ > -- Born in Roswell... married an alien... http://emptyhammock.com/
Re: Does mod_perl/mod_??? need a hook called when a request/conn leaves the original worker thread?
On Nov 18, 2013, at 12:15 PM, Eric Covener wrote: > On Mon, Nov 18, 2013 at 10:58 AM, Jeff Trawick wrote: >> For the mod_perl crash with Event that I posted at the URL below, I would >> suspect that there's some affinity with the original worker thread. Can >> anyone in mod_perl land confirm? >> >> http://mail-archives.apache.org/mod_mbox/perl-dev/201311.mbox/%3CCAKUrXK6C3R_F3NdA%2BJUGYOqppvnoQJLTGQ9%2BA916vuMb0g9dig%40mail.gmail.com%3E >> >> I'm also aware of a third-party diagnostic module that could use a hint, and >> in general I wonder if anyone knows of specific interface requirements that >> would need to be provided by a new hook for indicating when a connection or >> request leaves the original thread or is handled by a new one. > > I can't see any need for more than request_rec on both ends. > mod_spdy uses a fake conn_rec, well, more like a sub-conn_rec ala a subrequest, which is also sometimes called a slave connection... I'm thinking that maybe having such an entity here would be best, long term.
Re: Does mod_perl/mod_??? need a hook called when a request/conn leaves the original worker thread?
On Mon, Nov 18, 2013 at 10:58 AM, Jeff Trawick wrote: > For the mod_perl crash with Event that I posted at the URL below, I would > suspect that there's some affinity with the original worker thread. Can > anyone in mod_perl land confirm? > > http://mail-archives.apache.org/mod_mbox/perl-dev/201311.mbox/%3CCAKUrXK6C3R_F3NdA%2BJUGYOqppvnoQJLTGQ9%2BA916vuMb0g9dig%40mail.gmail.com%3E > > I'm also aware of a third-party diagnostic module that could use a hint, and > in general I wonder if anyone knows of specific interface requirements that > would need to be provided by a new hook for indicating when a connection or > request leaves the original thread or is handled by a new one. I can't see any need for more than request_rec on both ends.
