[Spread-users] Apache Logs sent using spread.

Torsten Curdt tcurdt at vafer.org
Wed Jun 4 08:09:42 EDT 2008

>     I have a sender application(similar to cronolog) which sends the  
> log messages from Apache to Spread daemon.

You mean cron-like ...not as they arrive?

> I have a receiver application which receives messages from the  
> spread daemon (Both communicate using a common group).
> I have few question regarding this.
> 1) Can you suggest a better application which can recieve log  
> messages from multiple Apache servers and log them into a central  
> repository.
>     It has to be reliable and fault tolerant like spread.

I am wondering why you are looking into an alternative ....because I  
am currently looking into setting something like that up for us.

>  2) I connect to a spread daemon running on a partcular system. I  
> would like to know, what happens when the spread daemon itself dies.

Well, apache will not be able to send the message to the ring. Up to  
you how you deal with that I guess.

>     Is there a way the communication still happen between the sender  
> and receiver through a different daemon automatically. If it can,  
> can you suggest how?

I think mod_log_spread has a fall back. But I couldn't get  
mod_log_spread to work properly ...and it does not support error logs  

So I have patched rotatelogs to push the logs into the spread ring.

I find the spread situation quite bad. No spread4 in debian/ubuntu. No  
easy to install mod_log_spread. Not that nice.
There are many things to improve (handling of config errors is  
terrible!) but I see no progress. Wondering about the project health  

>  3) In a spread_segment of the configuration we have multiple IP  
> addresses of different daemons.
>     ex:-  spread_segment {
>                                     machineA
>                                     machineB
>                                     machineC
>             }
>     However, A sender through the API's has to explicitly connect to  
> a single daemon on a system. I am curious as to why we have multiple  
> systems in a spread segment.

You mean why the whole ring needs to be defined/configured on each an  
every machine participating?


More information about the Spread-users mailing list