httpd-modules-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Vaughan Reid" <i...@cubos.com.au>
Subject RE: MySQL Virtual Host and Traffic Module
Date Thu, 19 Mar 2009 01:12:07 GMT
Hi Dave,

Thanks for the response. I haven't thought of doing the SQL query the way
you suggested, however I agree that it will cause unnecessary load on busy
servers and I would like to keep this as efficient as possible.

The second option sounds more reasonable. I have already used threading to
make a function which ticks on a configurable interval so I suppose each
child process would dump data for each of its vhosts at this interval, using
a query similar to what you have suggested.

I think I might go with the second option for the time being and see how it
goes but I am still interested to know if there is a way to store per vhost
data across children?


Thanks,
Vaughan

-----Original Message-----
From: Dave Ingram [mailto:dave@dmi.me.uk] 
Sent: Thursday, 19 March 2009 12:28 AM
To: modules-dev@httpd.apache.org
Subject: Re: MySQL Virtual Host and Traffic Module

Vaughan,

> What I have so far are 2 filters which gather the inbound traffic and
> outbound traffic for each transaction. These work ok and when logging
> transactions to file all of the in/out byte amounts appear to be correct.
> The first problem however, is that each child has its own set of memory
and
> therefore keeps its own totals per virtual host. This also means that
> multiple logging events occur for each transaction. I could just log this
> all to database but it would 1) be inefficient and 2) cause the size of
the
> database to grow quite quickly.
>   

It sounds to me like you could go two ways with this. I don't know the
format of your database table, but it should be possible to update it
atomically using something like:

INSERT INTO bandwidth (vhost_id, bw_in, bw_out) VALUES (42, 1124,
5023409) ON DUPLICATE KEY UPDATE bw_in = bw_in + 1124, bw_out = bw_out +
5023409

but that could lead to a lot of load. Another way might be for each
child to collect statistics and only flush to the database periodically,
say every 30 seconds (perhaps configurable on a per-vhost basis, so that
load-heavy sites could have larger update intervals). It would still be
possible to use the query above though.

This query could probably even be updated to split statistics on a
date/time basis, if you require more granular reporting.

Or have I missed/misunderstood something?


Dave


Mime
View raw message