Re: Fwd: [PHP] php with threaded MPM problem ?
| From: | jvlad | Date: | Wed, 19 Dec 2007 23:39:23 +0000 |
| Subject: | Re: Fwd: [PHP] php with threaded MPM problem ? | ||
| References: | 1 2 3 4 5 6 | Groups: | php.internals |
| Request: | Send a blank email to internals+get-34111@lists.php.net to get a copy of this message | ||
I use graceful restart quite often.
Frankly, I have no idea why it may fail to work :)
Indeed, when you request a graceful restart, apache will start next set of
children processes and all existing processes will stop getting and
processing any new requests so they will exit as soon as they finish current
ones.
It's a very neat feature, especially if you care of your visitors and for
example don't want to break their long downloads.
Finally it works fine for me (apache 2.0 + php 4.3)
If you get any problems I'd recommend to check extended status (see Apache
manual on what it is) and enable debug level for the logs.
In case of crashes in php, please follow appropriate instructions and submit
report at bugs.php.net
"Jochem Maas" <jochem@iamjochem.com> wrote in message
news:47693F6A.2050808@iamjochem.com...
> Brian Moon schreef:
>> Rashmi Badan wrote:
>>> The test was trying to load mod_php between graceful restarts, i.e start
>>> apache without mod_php, then modify the config file to load it and do a
>>> graceful restart. This clearly fails because mod_php loads only on the
>>> second load within apache. It does this by setting some pool user data
>>> in
>>> the beginning of sapi_apache2.c:php_apache_server_startup(). So I'd
>>> have to
>>> do two graceful restarts for it to take effect.
>>
>> PHP has never worked well with graceful restarts. Frankly, there are
>> some Apache things that never seemed to do well either. I gave up on
>> graceful restarts 8 years ago. Maybe they are better now.
>
> could anyone else confirm/deny that this is [still] the case?
> (i.e. avoid graceful restarts when using php)
>
>>
>> Just do a full restart and all will be fine. If this is so mission
>> critical that you can't be down for 2 seconds, you need more servers and
>> load balancing. Or if you are restarting your server that often,
>> something else is likely wrong.
>>
>> As for changing the code, a patch that does not break anything would be
>> the way to approach the list. It was probably done the way it is done
>> for a reason.
>>