[horde] WG: General question about clustering horde

[Rick Romero]

  Quoting Patrick De Zordo <patrick at spamreducer.eu>:

> -----Ursprüngliche Nachricht-----
> Von: Patrick De Zordo [mailto:patrick at spamreducer.eu]
> Gesendet: Donnerstag, 5. Juni 2014 16:18
> An: 'Ralf Lang'
> Betreff: AW: [horde] General question about clustering horde
>
>> -----Ursprüngliche Nachricht-----
>> Von: horde-bounces at lists.horde.org [mailto:horde-
>> bounces at lists.horde.org] Im Auftrag von Ralf Lang
>> Gesendet: Donnerstag, 5. Juni 2014 12:47
>> An: horde at lists.horde.org
>> Betreff: Re: [horde] General question about clustering horde
>>
>> On 05.06.2014 10:09, Patrick De Zordo wrote:
>> Hi guys,
>>
>> we are going on with testing this great state of the art product.
>>
>> Now we are testing to cluster 2 horde servers and would like to know
>> if there are some “sticking points” we should considering when
>> trying to cluster Horde.
>>
>> We don’t even know if someone has done this before..?
>>
>> Thank for giving me your attention and informations!
>>
>> If you don't share session backends, make sessions sticky through an
>> apache cookie on the load balancer or siminal.
>>
>> Otherwise user may experience occassional logouts.
>
> Since our horde session backend is a SQL DB it should be possible to run
> this as master/master synchronized DB, hopefully that would/could work..
> Anyway, thanks for the hint, will be added to the actual To-Do list..!

Also, I'm not sure where you want to store your temp files or how your load
balancing will work. There are temp files from viewers (like wv, excel,
etc), and temp upload files. I don't like putting uploads in SQL, and I
don't think the viewer temp files would be put into SQL anyways - so I use
glusterfs for my php & horde local temp directory and replicate that to
my  other web server. 

Then if the first server fails, your load balancing isn't "sticky", or the
remote user's IP varies, users won't have issues with some of those temp
files 'disappearing'.

Rick