[horde] ActiveSync problem

[Simon Wilson]

Quoting Simon Wilson <simon at simonandkate.net>:

> Quoting Simon Wilson <simon at simonandkate.net>:
>
>> Quoting Patrick Domack <patrickdk at patrickdk.com>:
>>
>>> Quoting Simon Wilson <simon at simonandkate.net>:
>>>
>>>> Quoting Simon Wilson <simon at simonandkate.net>:
>>>>
>>>>> On 18/07/2011, at 1:00 AM, Michael J Rubinsky <mrubinsk at horde.org> wrote:
>>>>>
>>>>>> Simon Wilson <simon at simonandkate.net> wrote:
>>>>>>
>>>>>>>
>>>>>>> I have been trying to get ActiveSync working for a while with limited
>>>>>>> success.
>>>>>>>
>>>>>>> Horde 4.0.7, Turba 3.0.4. Hordeauth is set to use LDAP. Turba links
>>>>>>> into LDAP for both a Personal LDAP address book and also a Shared
>>>>>>> Address Book.
>>>>>>>
>>>>>>> <snip>
>>>>>>>
>>>>>>> Initial discovery and sync - works fine. The iPad "sees" Horde, and
>>>>>>> syncs the Contacts from LDAP immediately.
>>>>>>> Changes in Turba - add, delete, edit: all sync almost immediately thru
>>>>>>> to the iPad.
>>>>>>> Changes on iPad end - This is the problem child. Nothing happens in
>>>>>>> Turba. The iPad sits spinning its "network activity" wheel, and Turba
>>>>>>> doesn't update. <snip>
>>>>>>
>>>>>> That sounds as if there are no more apace or php-cgi processes  
>>>>>> available to answer the HTTP requests, i.e., a configuration  
>>>>>> issue.
>>>>>> --
>>>>>> Mike
>>>>>> Sent from mobile
>>>>>>
>>>>> Hmmm... Thanks Mike. I have been running Apache with a reduced  
>>>>> spareserver count and a faster server turnover as the Apache  
>>>>> instances tend to otherwise take over the server. Horde has been  
>>>>> running fine, but I will reset to out of box defaults and see  
>>>>> how it goes.
>>>>>
>>>>> Cheers
>>>>> Simon
>>>>> --
>>>>
>>>> OK, put Apache back to stock config for the prefork MPM -
>>>>
>>>> StartServers 8
>>>> MinSpareServers 5
>>>> MaxSpareServes 20
>>>> ServerLimit 256
>>>> MaxClients 256
>>>> MaxRequestsPerChild 4000
>>>>
>>>> Note that at most I have 4 concurrent users.
>>>>
>>>> <snip>
>>>>
>>>> This is a bog standard CentOS server with "yum install httpd",  
>>>> and the config is now as per out of box. If I need to boost  
>>>> Apache servers beyond standard to run Horde ActiveSync then some  
>>>> sort of guide as to "what and by how much" would be great. How  
>>>> does everyone else have their Apache configured?
>>>>
>>>> Boosting Apache beyond these will also require more RAM - the  
>>>> server has 1.5GB RAM, and ONLY runs Apache / PHP / Horde. Surely  
>>>> that's enough for Horde?
>>>
>>> That seems odd, horde should run just find in 1gig of ram, though  
>>> I wouldn't if you have a lot of users using it. Currently my  
>>> personal email (only has about 10 users on it) has 2gigs of ram,  
>>> running everything  
>>> (postfix+dovecot+amavis+spamassassin+clamav+apache+php+ftp+mysql+quagga+...)
>>>
>>>   StartServers          8
>>>   MinSpareServers       8
>>>   MaxSpareServers      10
>>>   MaxClients           50
>>>   MaxRequestsPerChild 1000
>>>
>>> currently apache is using about 200megs of ram.
>>>
>>> I would only load the min required php modules needed though, some  
>>> of them have nasty effects. I know for me, gd leaks memory, and  
>>> smnp just adds some kind of huge delay to each script start.
>>>
>>>
>>
>> Thanks Patrick, will check what is being loaded. I haven't changed  
>> PHP config other than adding modules needed by Horde and I had to  
>> change memory to 128MB (was running fine on 96MB, but address book  
>> import falls over).
>>
>>
>
> I have a separate virtual running postfix, cyrus, etc. This one is  
> only Apache / PHP / Horde (with mysql for prefs etc).
>
> PHP is loading the following in php.ini:
>
> geoip idn lzf fileinfo imagick apc
>
> And the php.d directory includes:
>
> dbase dom gd imap json ldap mbstring mcrypt mhash mysql pdo  
> pdo_mysql pdo_sqlite tidy xmlreader xmlwriter xsl zip
>
> I need most of those I believe. Happy to leave some out if they  
> aren't needed. With that and the httpd defaults as noted above I  
> currently have 24 httpd processes running, each with about 75M RES  
> memory, i.e. more than I have.
>
> Note that the server has run fine with 5 apache servers started and  
> 10 max spare for 3 months with Horde4 without ActiveSync well within  
> the 1.5GB RAM. From the memory it is now chewing I am going to have  
> to pull back either the number of spareservers again, or work out  
> how to reduce httpd's memory usage.
>
> All in all if it is Apache running out of resources causing 25  
> minute sync delays then I can't see how I can run ActiveSync using  
> this CentOS server's default Apache setup on 1.5GB RAM.
>
> Simon

I have added some RAM, the server now has 2GB. Apache config is  
default MPM Prefork config:

StartServers 8
MinSpareServers 5
MaxSpareServers 20
ServerLimit 256
MaxClients 256
MaxRequestsPerChild 4000

and there are LOTS of free servers. Horde response is quick, and  
remains so throughout sync issues.

Sync remains reliable in one direction only - server to client. I  
tried again, edited a contact on my iPhone, and it took an hour before  
it replicated to Turba. Another try, and before that one had synced,  
made another one to the same field. Only the original one replicated,  
in an hour, the final edit was lost completely.

Restart Apache and the syncs go through immediately. At no stage does  
Apache behave like it is running out of servers or resources, nor is  
it logging any errors.

Simon.