FYI - I just checked again, and it jumped:

accepted conn: 1372 pool: www process manager: static idle processes: 98
active processes: 2 total processes: 100

I guess it took a second for the status to update.


On Fri, Oct 8, 2010 at 2:54 PM, Ilan Berkner <iberkner@gmail.com> wrote:

> Ok, I got status working:
>
> accepted conn: 72 pool: www process manager: static idle processes: 99
> active processes: 1 total processes: 100
>
> which data points matter to me from the above information?
>
> On Fri, Oct 8, 2010 at 2:42 PM, Ilan Berkner <iberkner@gmail.com> wrote:
>
>> At this time, I do not monitor php-fpm via its status page, excellent
>> suggestion, I'll see what I can do to turn that on.
>>
>> I submitted the bug: 53028
>>
>> What's a good article on how to enable the php-fpm status page via Nginx,
>> I remember reading about it somewhere.
>>
>> Thanks
>>
>> 2010/10/8 Jérôme Loyet <ml@fatbsd.com>
>>
>>> 2010/10/8 Ilan Berkner <iberkner@gmail.com>:
>>> > I understand what you mean, in terms of that its not often and it does
>>> > appear to be working...
>>> > Am happy to open a bug report, how / where do I open a bug report and
>>> what
>>> > information is required?
>>>
>>> Report here http://bugs.php.net/report.php but read first
>>> http://bugs.php.net/how-to-report.php
>>>
>>> > Right now I'm running as static with max children set to 100. Do you
>>> > recommend switching to dynamic, if so, what would be good settings for
>>> a
>>> > busy site? Each box has 4GB and strictly runs php-fpm, can add more
>>> RAM if
>>> > needed, but there seems to be between 500MB and 1GB free at all times
>>> on
>>> > theses systems, so I think I'm OK with RAM for now.
>>>
>>> do you monitor PHP-FPM activity through its status page ? Look how
>>> php-fpm is used in order to determine if max_children is set right.
>>> After having several days of data you'll be able to tweak and see how
>>> it goes ... and start again ... :)
>>>
>>> >
>>> > 2010/10/8 Jérôme Loyet <ml@fatbsd.com>
>>> >>
>>> >> 2010/10/8 Ilan Berkner <iberkner@gmail.com>:
>>> >> > FYI - I am seeing this every so often:
>>> >> > Oct 08 08:48:14.087065 [ERROR] pid 1549,
>>> >> > fpm_pctl_perform_idle_server_maintenance(), line 347: [pool www]
>>> unable
>>> >> > to
>>> >> > retrieve process activity of one or more child(ren). Will try again
>>> >> > later.
>>> >> > Oct 08 10:22:01.829260 [ERROR] pid 1549,
>>> >> > fpm_pctl_perform_idle_server_maintenance(), line 347: [pool www]
>>> unable
>>> >> > to
>>> >> > retrieve process activity of one or more child(ren). Will try again
>>> >> > later.
>>> >> > Oct 08 10:46:36.296591 [ERROR] pid 1549,
>>> >> > fpm_pctl_perform_idle_server_maintenance(), line 347: [pool www]
>>> unable
>>> >> > to
>>> >> > retrieve process activity of one or more child(ren). Will try again
>>> >> > later.
>>> >> > Oct 08 11:53:41.484788 [ERROR] pid 1549,
>>> >> > fpm_pctl_perform_idle_server_maintenance(), line 347: [pool www]
>>> unable
>>> >> > to
>>> >> > retrieve process activity of one or more child(ren). Will try again
>>> >> > later.
>>> >> > Oct 08 12:35:35.574208 [ERROR] pid 1549,
>>> >> > fpm_pctl_perform_idle_server_maintenance(), line 347: [pool www]
>>> unable
>>> >> > to
>>> >> > retrieve process activity of one or more child(ren). Will try again
>>> >> > later.
>>> >> >
>>> >>
>>> >> it's not so often, this function is called every seconds and you have
>>> >> errors with several minutes between each. But you should open a bug
>>> >> report so that we look for a correction.
>>> >>
>>> >> > 2010/10/7 Jérôme Loyet <ml@fatbsd.com>
>>> >> >>
>>> >> >> 2010/10/7 Ilan Berkner <iberkner@gmail.com>:
>>> >> >> > Could someone make sure I understand, debug will give me the
>>> highest
>>> >> >> > level
>>> >> >> > of error information, correct? this is what I'm looking for.
>>> >> >> > I just got this error:
>>> >> >> >
>>> >> >> > Oct 07 12:41:38.287422 [ERROR] pid 27916,
>>> >> >> > fpm_pctl_perform_idle_server_maintenance(), line 347: [pool www]
>>> >> >> > unable
>>> >> >> > to
>>> >> >> > retrieve process activity of one or more child(ren). Will try
>>> again
>>> >> >> > later.
>>> >> >> >
>>> >> >> > What does that mean??
>>> >> >>
>>> >> >> it means that the parent process has not been able to retrieve
>>> >> >> consistent informations about its children (if each child is idle
>>> or
>>> >> >> active and if idle + active = total_children).
>>> >> >> If the static PM is used, it just means that for once the status
>>> page
>>> >> >> won't be up to date, so there is nothing to care about.
>>> >> >> If the dynamic PM is used, it means that the parent won't do
>>> anything
>>> >> >> until it has consistent informations.
>>> >> >>
>>> >> >> So if the errors triggers once this not a problem. If it happens
>>> every
>>> >> >> second (the frequency of
>>> fpm_pctl_perform_idle_server_maintenance())
>>> >> >> it might be a problem.
>>> >> >>
>>> >> >> >
>>> >> >> > On Thu, Oct 7, 2010 at 1:28 PM, Jonathan Langevin
>>> >> >> > <intel352@gmail.com>
>>> >> >> > wrote:
>>> >> >> >>
>>> >> >> >> I thought notice was inclusive of debug, not the other way
>>> around?
>>> >> >> >> Debug is a higher level of log output, whereas notice is not.
>>> >> >> >>
>>> >> >> >> 2010/10/7 Jérôme Loyet <ml@fatbsd.com>
>>> >> >> >>>
>>> >> >> >>> 2010/10/7 Ilan Berkner <iberkner@gmail.com>:
>>> >> >> >>> > Why would notice be more appropriate? Right now I'm trying
>>> to
>>> >> >> >>> > get
>>> >> >> >>> > as
>>> >> >> >>> > much
>>> >> >> >>> > information out of the system as possible. Is "debug"
>>> inclusive
>>> >> >> >>> > of
>>> >> >> >>> > notice?
>>> >> >> >>>
>>> >> >> >>> yes
>>> >> >> >>>
>>> >> >> >>> >
>>> >> >> >>> > 2010/10/7 Jérôme Loyet <ml@fatbsd.com>
>>> >> >> >>> >>
>>> >> >> >>> >> 2010/10/7 Ilan Berkner <iberkner@gmail.com>:
>>> >> >> >>> >> > Thanks, I'm still trying to understand what the bottleneck
>>> is.
>>> >> >> >>> >> > What is the highest error reporting level: debug?
>>> >> >> >>> >>
>>> >> >> >>> >> yes but notice should be more appropriate
>>> >> >> >>> >>
>>> >> >> >>> >>
>>> >> >> >>> >> > Thanks
>>> >> >> >>> >> >
>>> >> >> >>> >> > 2010/10/7 Jérôme Loyet <ml@fatbsd.com>
>>> >> >> >>> >> >>
>>> >> >> >>> >> >> 2010/10/7 Ilan Berkner <iberkner@gmail.com>:
>>> >> >> >>> >> >> > Hi All,
>>> >> >> >>> >> >> > Looking for some general guidance on php-fpm settings
>>> for a
>>> >> >> >>> >> >> > very
>>> >> >> >>> >> >> > busy
>>> >> >> >>> >> >> > site
>>> >> >> >>> >> >> > 3+mm page views a day.
>>> >> >> >>> >> >> > We have 4 php app servers and they are getting killed
>>> >> >> >>> >> >> > during
>>> >> >> >>> >> >> > busy
>>> >> >> >>> >> >> > time,
>>> >> >> >>> >> >> > I'm
>>> >> >> >>> >> >> > going to add 2 more, but started looking at php-fpm
>>> logs
>>> >> >> >>> >> >> > and
>>> >> >> >>> >> >> > seeing
>>> >> >> >>> >> >> > things
>>> >> >> >>> >> >> > about warnings that php-fpm is busy, that I should add
>>> more
>>> >> >> >>> >> >> > min/max /
>>> >> >> >>> >> >> > start
>>> >> >> >>> >> >> > servers, so I'm sure our configuration can be improved.
>>> >> >> >>> >> >> > Any suggestions as to:
>>> >> >> >>> >> >> > 1) use static or dynamic?
>>> >> >> >>> >> >>
>>> >> >> >>> >> >> depends
>>> >> >> >>> >> >>
>>> >> >> >>> >> >> > 2) If static, what should max children be set to?
>>> >> >> >>> >> >>
>>> >> >> >>> >> >> 42 ?
>>> >> >> >>> >> >>
>>> >> >> >>> >> >> > 3) if dynamic, what are ideal parameters for the
>>> dynamic
>>> >> >> >>> >> >> > stuff
>>> >> >> >>> >> >> > for
>>> >> >> >>> >> >> > very
>>> >> >> >>> >> >> > busy
>>> >> >> >>> >> >> > situations?
>>> >> >> >>> >> >>
>>> >> >> >>> >> >> More seriously it's complicated to find the best conf. If
>>> >> >> >>> >> >> it's a
>>> >> >> >>> >> >> very
>>> >> >> >>> >> >> busy site, I recommand to :
>>> >> >> >>> >> >> 1- 1 or 2 servers for running nginx or any other reverse
>>> >> >> >>> >> >> proxy
>>> >> >> >>> >> >> 2- several servers to run php-fpm and only php-fpm.
>>> >> >> >>> >> >> 3- use static or dynamic with max_spare_children =
>>> >> >> >>> >> >> max_children
>>> >> >> >>> >> >> (you
>>> >> >> >>> >> >> want to be ready to handle the heavy load).
>>> >> >> >>> >> >> 4- set max_children to something which sounds great to
>>> you at
>>> >> >> >>> >> >> the
>>> >> >> >>> >> >> first place and adjust after depending on the load and
>>> the
>>> >> >> >>> >> >> usage
>>> >> >> >>> >> >> of
>>> >> >> >>> >> >> the server. If you have 4GB of RAM and each php-fpm
>>> process
>>> >> >> >>> >> >> uses
>>> >> >> >>> >> >> on
>>> >> >> >>> >> >> average 32MB you can set max children to 4096/32 - 10%.
>>> But
>>> >> >> >>> >> >> you
>>> >> >> >>> >> >> will
>>> >> >> >>> >> >> never find the right and perfect solution at the first
>>> try.
>>> >> >> >>> >> >>
>>> >> >> >>> >> >> > Thanks
>>> >> >> >>> >> >
>>> >> >> >>> >> >
>>> >> >> >>> >
>>> >> >> >>> >
>>> >> >> >>
>>> >> >> >
>>> >> >> >
>>> >> >
>>> >> >
>>> >
>>> >
>>>
>>
>>
>

2010/10/8 Ilan Berkner <iberkner@gmail.com>:
> Ok, I got status working:

Can you please try not to spam this mailing liste each time you hit a key ?

> accepted conn: 72 pool: www process manager: static idle processes: 99
> active processes: 1 total processes: 100
> which data points matter to me from the above information?

Monitor "active processes"

>
> On Fri, Oct 8, 2010 at 2:42 PM, Ilan Berkner <iberkner@gmail.com> wrote:
>>
>> At this time, I do not monitor php-fpm via its status page, excellent
>> suggestion, I'll see what I can do to turn that on.
>> I submitted the bug: 53028
>> What's a good article on how to enable the php-fpm status page via Nginx,
>> I remember reading about it somewhere.
>> Thanks
>> 2010/10/8 Jérôme Loyet <ml@fatbsd.com>
>>>
>>> 2010/10/8 Ilan Berkner <iberkner@gmail.com>:
>>> > I understand what you mean, in terms of that its not often and it does
>>> > appear to be working...
>>> > Am happy to open a bug report, how / where do I open a bug report and
>>> > what
>>> > information is required?
>>>
>>> Report here http://bugs.php.net/report.php but read first
>>> http://bugs.php.net/how-to-report.php
>>>
>>> > Right now I'm running as static with max children set to 100.  Do you
>>> > recommend switching to dynamic, if so, what would be good settings for
>>> > a
>>> > busy site?  Each box has 4GB and strictly runs php-fpm, can add more
>>> > RAM if
>>> > needed, but there seems to be between 500MB and 1GB free at all times
>>> > on
>>> > theses systems, so I think I'm OK with RAM for now.
>>>
>>> do you monitor PHP-FPM activity through its status page ? Look how
>>> php-fpm is used in order to determine if max_children is set right.
>>> After having several days of data you'll be able to tweak and see how
>>> it goes ... and start again ... :)
>>>
>>> >
>>> > 2010/10/8 Jérôme Loyet <ml@fatbsd.com>
>>> >>
>>> >> 2010/10/8 Ilan Berkner <iberkner@gmail.com>:
>>> >> > FYI - I am seeing this every so often:
>>> >> > Oct 08 08:48:14.087065 [ERROR] pid 1549,
>>> >> > fpm_pctl_perform_idle_server_maintenance(), line 347: [pool www]
>>> >> > unable
>>> >> > to
>>> >> > retrieve process activity of one or more child(ren). Will try again
>>> >> > later.
>>> >> > Oct 08 10:22:01.829260 [ERROR] pid 1549,
>>> >> > fpm_pctl_perform_idle_server_maintenance(), line 347: [pool www]
>>> >> > unable
>>> >> > to
>>> >> > retrieve process activity of one or more child(ren). Will try again
>>> >> > later.
>>> >> > Oct 08 10:46:36.296591 [ERROR] pid 1549,
>>> >> > fpm_pctl_perform_idle_server_maintenance(), line 347: [pool www]
>>> >> > unable
>>> >> > to
>>> >> > retrieve process activity of one or more child(ren). Will try again
>>> >> > later.
>>> >> > Oct 08 11:53:41.484788 [ERROR] pid 1549,
>>> >> > fpm_pctl_perform_idle_server_maintenance(), line 347: [pool www]
>>> >> > unable
>>> >> > to
>>> >> > retrieve process activity of one or more child(ren). Will try again
>>> >> > later.
>>> >> > Oct 08 12:35:35.574208 [ERROR] pid 1549,
>>> >> > fpm_pctl_perform_idle_server_maintenance(), line 347: [pool www]
>>> >> > unable
>>> >> > to
>>> >> > retrieve process activity of one or more child(ren). Will try again
>>> >> > later.
>>> >> >
>>> >>
>>> >> it's not so often, this function is called every seconds and you have
>>> >> errors with several minutes between each. But you should open a bug
>>> >> report so that we look for a correction.
>>> >>
>>> >> > 2010/10/7 Jérôme Loyet <ml@fatbsd.com>
>>> >> >>
>>> >> >> 2010/10/7 Ilan Berkner <iberkner@gmail.com>:
>>> >> >> > Could someone make sure I understand, debug will give me the
>>> >> >> > highest
>>> >> >> > level
>>> >> >> > of error information, correct? this is what I'm looking for.
>>> >> >> > I just got this error:
>>> >> >> >
>>> >> >> > Oct 07 12:41:38.287422 [ERROR] pid 27916,
>>> >> >> > fpm_pctl_perform_idle_server_maintenance(), line 347: [pool www]
>>> >> >> > unable
>>> >> >> > to
>>> >> >> > retrieve process activity of one or more child(ren). Will try
>>> >> >> > again
>>> >> >> > later.
>>> >> >> >
>>> >> >> > What does that mean??
>>> >> >>
>>> >> >> it means that the parent process has not been able to retrieve
>>> >> >> consistent informations about its children (if each child is idle
>>> >> >> or
>>> >> >> active and if idle + active = total_children).
>>> >> >> If the static PM is used, it just means that for once the status
>>> >> >> page
>>> >> >> won't be up to date, so there is nothing to care about.
>>> >> >> If the dynamic PM is used, it means that the parent won't do
>>> >> >> anything
>>> >> >> until it has consistent informations.
>>> >> >>
>>> >> >> So if the errors triggers once this not a problem. If it happens
>>> >> >> every
>>> >> >> second (the frequency of
>>> >> >> fpm_pctl_perform_idle_server_maintenance())
>>> >> >> it might be a problem.
>>> >> >>
>>> >> >> >
>>> >> >> > On Thu, Oct 7, 2010 at 1:28 PM, Jonathan Langevin
>>> >> >> > <intel352@gmail.com>
>>> >> >> > wrote:
>>> >> >> >>
>>> >> >> >> I thought notice was inclusive of debug, not the other way
>>> >> >> >> around?
>>> >> >> >> Debug is a higher level of log output, whereas notice is not.
>>> >> >> >>
>>> >> >> >> 2010/10/7 Jérôme Loyet <ml@fatbsd.com>
>>> >> >> >>>
>>> >> >> >>> 2010/10/7 Ilan Berkner <iberkner@gmail.com>:
>>> >> >> >>> > Why would notice be more appropriate?  Right now I'm trying
>>> >> >> >>> > to
>>> >> >> >>> > get
>>> >> >> >>> > as
>>> >> >> >>> > much
>>> >> >> >>> > information out of the system as possible.  Is "debug"
>>> >> >> >>> > inclusive
>>> >> >> >>> > of
>>> >> >> >>> > notice?
>>> >> >> >>>
>>> >> >> >>> yes
>>> >> >> >>>
>>> >> >> >>> >
>>> >> >> >>> > 2010/10/7 Jérôme Loyet <ml@fatbsd.com>
>>> >> >> >>> >>
>>> >> >> >>> >> 2010/10/7 Ilan Berkner <iberkner@gmail.com>:
>>> >> >> >>> >> > Thanks, I'm still trying to understand what the bottleneck
>>> >> >> >>> >> > is.
>>> >> >> >>> >> > What is the highest error reporting level: debug?
>>> >> >> >>> >>
>>> >> >> >>> >> yes but notice should be more appropriate
>>> >> >> >>> >>
>>> >> >> >>> >>
>>> >> >> >>> >> > Thanks
>>> >> >> >>> >> >
>>> >> >> >>> >> > 2010/10/7 Jérôme Loyet <ml@fatbsd.com>
>>> >> >> >>> >> >>
>>> >> >> >>> >> >> 2010/10/7 Ilan Berkner <iberkner@gmail.com>:
>>> >> >> >>> >> >> > Hi All,
>>> >> >> >>> >> >> > Looking for some general guidance on php-fpm settings
>>> >> >> >>> >> >> > for a
>>> >> >> >>> >> >> > very
>>> >> >> >>> >> >> > busy
>>> >> >> >>> >> >> > site
>>> >> >> >>> >> >> > 3+mm page views a day.
>>> >> >> >>> >> >> > We have 4 php app servers and they are getting killed
>>> >> >> >>> >> >> > during
>>> >> >> >>> >> >> > busy
>>> >> >> >>> >> >> > time,
>>> >> >> >>> >> >> > I'm
>>> >> >> >>> >> >> > going to add 2 more, but started looking at php-fpm
>>> >> >> >>> >> >> > logs
>>> >> >> >>> >> >> > and
>>> >> >> >>> >> >> > seeing
>>> >> >> >>> >> >> > things
>>> >> >> >>> >> >> > about warnings that php-fpm is busy, that I should add
>>> >> >> >>> >> >> > more
>>> >> >> >>> >> >> > min/max /
>>> >> >> >>> >> >> > start
>>> >> >> >>> >> >> > servers, so I'm sure our configuration can be improved.
>>> >> >> >>> >> >> > Any suggestions as to:
>>> >> >> >>> >> >> > 1) use static or dynamic?
>>> >> >> >>> >> >>
>>> >> >> >>> >> >> depends
>>> >> >> >>> >> >>
>>> >> >> >>> >> >> > 2) If static, what should max children be set to?
>>> >> >> >>> >> >>
>>> >> >> >>> >> >> 42 ?
>>> >> >> >>> >> >>
>>> >> >> >>> >> >> > 3) if dynamic, what are ideal parameters for the
>>> >> >> >>> >> >> > dynamic
>>> >> >> >>> >> >> > stuff
>>> >> >> >>> >> >> > for
>>> >> >> >>> >> >> > very
>>> >> >> >>> >> >> > busy
>>> >> >> >>> >> >> > situations?
>>> >> >> >>> >> >>
>>> >> >> >>> >> >> More seriously it's complicated to find the best conf. If
>>> >> >> >>> >> >> it's a
>>> >> >> >>> >> >> very
>>> >> >> >>> >> >> busy site, I recommand to :
>>> >> >> >>> >> >> 1- 1 or 2 servers for running nginx or any other reverse
>>> >> >> >>> >> >> proxy
>>> >> >> >>> >> >> 2- several servers to run php-fpm and only php-fpm.
>>> >> >> >>> >> >> 3- use static or dynamic with max_spare_children =
>>> >> >> >>> >> >> max_children
>>> >> >> >>> >> >> (you
>>> >> >> >>> >> >> want to be ready to handle the heavy load).
>>> >> >> >>> >> >> 4- set max_children to something which sounds great to
>>> >> >> >>> >> >> you at
>>> >> >> >>> >> >> the
>>> >> >> >>> >> >> first place and adjust after depending on the load and
>>> >> >> >>> >> >> the
>>> >> >> >>> >> >> usage
>>> >> >> >>> >> >> of
>>> >> >> >>> >> >> the server. If you have 4GB of RAM and each php-fpm
>>> >> >> >>> >> >> process
>>> >> >> >>> >> >> uses
>>> >> >> >>> >> >> on
>>> >> >> >>> >> >> average 32MB you can set max children to 4096/32 - 10%.
>>> >> >> >>> >> >> But
>>> >> >> >>> >> >> you
>>> >> >> >>> >> >> will
>>> >> >> >>> >> >> never find the right and perfect solution at the first
>>> >> >> >>> >> >> try.
>>> >> >> >>> >> >>
>>> >> >> >>> >> >> > Thanks
>>> >> >> >>> >> >
>>> >> >> >>> >> >
>>> >> >> >>> >
>>> >> >> >>> >
>>> >> >> >>
>>> >> >> >
>>> >> >> >
>>> >> >
>>> >> >
>>> >
>>> >
>>
>
>

No problem, thanks for all your suggestions, trying to get better informed.

In terms of active processes, what am I looking for? so far I've seen it
fluctuate between 0 and 5 at most, although right now is not our peak time.

Are you aware of any Cacti templates for graphing this data historically?

Thanks

2010/10/8 Jérôme Loyet <ml@fatbsd.com>

> 2010/10/8 Ilan Berkner <iberkner@gmail.com>:
> > Ok, I got status working:
>
> Can you please try not to spam this mailing liste each time you hit a key ?
>
> > accepted conn: 72 pool: www process manager: static idle processes: 99
> > active processes: 1 total processes: 100
> > which data points matter to me from the above information?
>
> Monitor "active processes"
>
> >
> > On Fri, Oct 8, 2010 at 2:42 PM, Ilan Berkner <iberkner@gmail.com> wrote:
> >>
> >> At this time, I do not monitor php-fpm via its status page, excellent
> >> suggestion, I'll see what I can do to turn that on.
> >> I submitted the bug: 53028
> >> What's a good article on how to enable the php-fpm status page via
> Nginx,
> >> I remember reading about it somewhere.
> >> Thanks
> >> 2010/10/8 Jérôme Loyet <ml@fatbsd.com>
> >>>
> >>> 2010/10/8 Ilan Berkner <iberkner@gmail.com>:
> >>> > I understand what you mean, in terms of that its not often and it
> does
> >>> > appear to be working...
> >>> > Am happy to open a bug report, how / where do I open a bug report and
> >>> > what
> >>> > information is required?
> >>>
> >>> Report here http://bugs.php.net/report.php but read first
> >>> http://bugs.php.net/how-to-report.php
> >>>
> >>> > Right now I'm running as static with max children set to 100. Do you
> >>> > recommend switching to dynamic, if so, what would be good settings
> for
> >>> > a
> >>> > busy site? Each box has 4GB and strictly runs php-fpm, can add more
> >>> > RAM if
> >>> > needed, but there seems to be between 500MB and 1GB free at all times
> >>> > on
> >>> > theses systems, so I think I'm OK with RAM for now.
> >>>
> >>> do you monitor PHP-FPM activity through its status page ? Look how
> >>> php-fpm is used in order to determine if max_children is set right.
> >>> After having several days of data you'll be able to tweak and see how
> >>> it goes ... and start again ... :)
> >>>
> >>> >
> >>> > 2010/10/8 Jérôme Loyet <ml@fatbsd.com>
> >>> >>
> >>> >> 2010/10/8 Ilan Berkner <iberkner@gmail.com>:
> >>> >> > FYI - I am seeing this every so often:
> >>> >> > Oct 08 08:48:14.087065 [ERROR] pid 1549,
> >>> >> > fpm_pctl_perform_idle_server_maintenance(), line 347: [pool www]
> >>> >> > unable
> >>> >> > to
> >>> >> > retrieve process activity of one or more child(ren). Will try
> again
> >>> >> > later.
> >>> >> > Oct 08 10:22:01.829260 [ERROR] pid 1549,
> >>> >> > fpm_pctl_perform_idle_server_maintenance(), line 347: [pool www]
> >>> >> > unable
> >>> >> > to
> >>> >> > retrieve process activity of one or more child(ren). Will try
> again
> >>> >> > later.
> >>> >> > Oct 08 10:46:36.296591 [ERROR] pid 1549,
> >>> >> > fpm_pctl_perform_idle_server_maintenance(), line 347: [pool www]
> >>> >> > unable
> >>> >> > to
> >>> >> > retrieve process activity of one or more child(ren). Will try
> again
> >>> >> > later.
> >>> >> > Oct 08 11:53:41.484788 [ERROR] pid 1549,
> >>> >> > fpm_pctl_perform_idle_server_maintenance(), line 347: [pool www]
> >>> >> > unable
> >>> >> > to
> >>> >> > retrieve process activity of one or more child(ren). Will try
> again
> >>> >> > later.
> >>> >> > Oct 08 12:35:35.574208 [ERROR] pid 1549,
> >>> >> > fpm_pctl_perform_idle_server_maintenance(), line 347: [pool www]
> >>> >> > unable
> >>> >> > to
> >>> >> > retrieve process activity of one or more child(ren). Will try
> again
> >>> >> > later.
> >>> >> >
> >>> >>
> >>> >> it's not so often, this function is called every seconds and you
> have
> >>> >> errors with several minutes between each. But you should open a bug
> >>> >> report so that we look for a correction.
> >>> >>
> >>> >> > 2010/10/7 Jérôme Loyet <ml@fatbsd.com>
> >>> >> >>
> >>> >> >> 2010/10/7 Ilan Berkner <iberkner@gmail.com>:
> >>> >> >> > Could someone make sure I understand, debug will give me the
> >>> >> >> > highest
> >>> >> >> > level
> >>> >> >> > of error information, correct? this is what I'm looking for.
> >>> >> >> > I just got this error:
> >>> >> >> >
> >>> >> >> > Oct 07 12:41:38.287422 [ERROR] pid 27916,
> >>> >> >> > fpm_pctl_perform_idle_server_maintenance(), line 347: [pool
> www]
> >>> >> >> > unable
> >>> >> >> > to
> >>> >> >> > retrieve process activity of one or more child(ren). Will try
> >>> >> >> > again
> >>> >> >> > later.
> >>> >> >> >
> >>> >> >> > What does that mean??
> >>> >> >>
> >>> >> >> it means that the parent process has not been able to retrieve
> >>> >> >> consistent informations about its children (if each child is idle
> >>> >> >> or
> >>> >> >> active and if idle + active = total_children).
> >>> >> >> If the static PM is used, it just means that for once the status
> >>> >> >> page
> >>> >> >> won't be up to date, so there is nothing to care about.
> >>> >> >> If the dynamic PM is used, it means that the parent won't do
> >>> >> >> anything
> >>> >> >> until it has consistent informations.
> >>> >> >>
> >>> >> >> So if the errors triggers once this not a problem. If it happens
> >>> >> >> every
> >>> >> >> second (the frequency of
> >>> >> >> fpm_pctl_perform_idle_server_maintenance())
> >>> >> >> it might be a problem.
> >>> >> >>
> >>> >> >> >
> >>> >> >> > On Thu, Oct 7, 2010 at 1:28 PM, Jonathan Langevin
> >>> >> >> > <intel352@gmail.com>
> >>> >> >> > wrote:
> >>> >> >> >>
> >>> >> >> >> I thought notice was inclusive of debug, not the other way
> >>> >> >> >> around?
> >>> >> >> >> Debug is a higher level of log output, whereas notice is not..
> >>> >> >> >>
> >>> >> >> >> 2010/10/7 Jérôme Loyet <ml@fatbsd.com>
> >>> >> >> >>>
> >>> >> >> >>> 2010/10/7 Ilan Berkner <iberkner@gmail.com>:
> >>> >> >> >>> > Why would notice be more appropriate? Right now I'm trying
> >>> >> >> >>> > to
> >>> >> >> >>> > get
> >>> >> >> >>> > as
> >>> >> >> >>> > much
> >>> >> >> >>> > information out of the system as possible. Is "debug"
> >>> >> >> >>> > inclusive
> >>> >> >> >>> > of
> >>> >> >> >>> > notice?
> >>> >> >> >>>
> >>> >> >> >>> yes
> >>> >> >> >>>
> >>> >> >> >>> >
> >>> >> >> >>> > 2010/10/7 Jérôme Loyet <ml@fatbsd.com>
> >>> >> >> >>> >>
> >>> >> >> >>> >> 2010/10/7 Ilan Berkner <iberkner@gmail.com>:
> >>> >> >> >>> >> > Thanks, I'm still trying to understand what the
> bottleneck
> >>> >> >> >>> >> > is.
> >>> >> >> >>> >> > What is the highest error reporting level: debug?
> >>> >> >> >>> >>
> >>> >> >> >>> >> yes but notice should be more appropriate
> >>> >> >> >>> >>
> >>> >> >> >>> >>
> >>> >> >> >>> >> > Thanks
> >>> >> >> >>> >> >
> >>> >> >> >>> >> > 2010/10/7 Jérôme Loyet <ml@fatbsd.com>
> >>> >> >> >>> >> >>
> >>> >> >> >>> >> >> 2010/10/7 Ilan Berkner <iberkner@gmail.com>:
> >>> >> >> >>> >> >> > Hi All,
> >>> >> >> >>> >> >> > Looking for some general guidance on php-fpm settings
> >>> >> >> >>> >> >> > for a
> >>> >> >> >>> >> >> > very
> >>> >> >> >>> >> >> > busy
> >>> >> >> >>> >> >> > site
> >>> >> >> >>> >> >> > 3+mm page views a day.
> >>> >> >> >>> >> >> > We have 4 php app servers and they are getting killed
> >>> >> >> >>> >> >> > during
> >>> >> >> >>> >> >> > busy
> >>> >> >> >>> >> >> > time,
> >>> >> >> >>> >> >> > I'm
> >>> >> >> >>> >> >> > going to add 2 more, but started looking at php-fpm
> >>> >> >> >>> >> >> > logs
> >>> >> >> >>> >> >> > and
> >>> >> >> >>> >> >> > seeing
> >>> >> >> >>> >> >> > things
> >>> >> >> >>> >> >> > about warnings that php-fpm is busy, that I should
> add
> >>> >> >> >>> >> >> > more
> >>> >> >> >>> >> >> > min/max /
> >>> >> >> >>> >> >> > start
> >>> >> >> >>> >> >> > servers, so I'm sure our configuration can be
> improved.
> >>> >> >> >>> >> >> > Any suggestions as to:
> >>> >> >> >>> >> >> > 1) use static or dynamic?
> >>> >> >> >>> >> >>
> >>> >> >> >>> >> >> depends
> >>> >> >> >>> >> >>
> >>> >> >> >>> >> >> > 2) If static, what should max children be set to?
> >>> >> >> >>> >> >>
> >>> >> >> >>> >> >> 42 ?
> >>> >> >> >>> >> >>
> >>> >> >> >>> >> >> > 3) if dynamic, what are ideal parameters for the
> >>> >> >> >>> >> >> > dynamic
> >>> >> >> >>> >> >> > stuff
> >>> >> >> >>> >> >> > for
> >>> >> >> >>> >> >> > very
> >>> >> >> >>> >> >> > busy
> >>> >> >> >>> >> >> > situations?
> >>> >> >> >>> >> >>
> >>> >> >> >>> >> >> More seriously it's complicated to find the best conf..
> If
> >>> >> >> >>> >> >> it's a
> >>> >> >> >>> >> >> very
> >>> >> >> >>> >> >> busy site, I recommand to :
> >>> >> >> >>> >> >> 1- 1 or 2 servers for running nginx or any other
> reverse
> >>> >> >> >>> >> >> proxy
> >>> >> >> >>> >> >> 2- several servers to run php-fpm and only php-fpm.
> >>> >> >> >>> >> >> 3- use static or dynamic with max_spare_children =
> >>> >> >> >>> >> >> max_children
> >>> >> >> >>> >> >> (you
> >>> >> >> >>> >> >> want to be ready to handle the heavy load).
> >>> >> >> >>> >> >> 4- set max_children to something which sounds great to
> >>> >> >> >>> >> >> you at
> >>> >> >> >>> >> >> the
> >>> >> >> >>> >> >> first place and adjust after depending on the load and
> >>> >> >> >>> >> >> the
> >>> >> >> >>> >> >> usage
> >>> >> >> >>> >> >> of
> >>> >> >> >>> >> >> the server. If you have 4GB of RAM and each php-fpm
> >>> >> >> >>> >> >> process
> >>> >> >> >>> >> >> uses
> >>> >> >> >>> >> >> on
> >>> >> >> >>> >> >> average 32MB you can set max children to 4096/32 - 10%.
> >>> >> >> >>> >> >> But
> >>> >> >> >>> >> >> you
> >>> >> >> >>> >> >> will
> >>> >> >> >>> >> >> never find the right and perfect solution at the first
> >>> >> >> >>> >> >> try.
> >>> >> >> >>> >> >>
> >>> >> >> >>> >> >> > Thanks
> >>> >> >> >>> >> >
> >>> >> >> >>> >> >
> >>> >> >> >>> >
> >>> >> >> >>> >
> >>> >> >> >>
> >>> >> >> >
> >>> >> >> >
> >>> >> >
> >>> >> >
> >>> >
> >>> >
> >>
> >
> >
>

would you mind sharing how you got status working ? :)