Correct me if I'm wrong but these options only show the stats regarding the
whole node, not per host...

What I'm trying to get is a *per-host/remap rule *"real-time" bandwidth
stats.
The plugin does aggregate bytes stats well, but it doesn't do that in
real-time. It seems that any solution to get this stats is going through
the log analyzing, but that solution is not accurate at all.


On Fri, Jul 19, 2013 at 4:48 PM, Mark Harrison <[email protected]> wrote:

> I'm not sure if it would suffer from the same problem or not, but you can
> try using the stats_over_http plugin or traffic_line -r and look at the 
> proxy.process.net.read_bytes
> and write_bytes variables.
>
>
> On Fri, Jul 19, 2013 at 8:52 AM, Ron Tsoref <[email protected]> wrote:
>
>> About half an year ago I looked for a solution to calculate the current
>> bandwidth served at any given time. I thought of using the logs to get this
>> information, but log lines are written only after a complete object was
>> served.  In the mailing list thread I opened back then, Conan posted a
>> plugin he wrote that show the bytes served for each host via a HTTP
>> interface and suggested me to take a look at it.
>>
>> The problem is that the bytes served for each host are increased only
>> after a complete object is served, just like what I can do with the
>> log-analyzing solution.
>>
>> This situation is causing spikes and bounces when analyzing the data from
>> this plugin in small intervals.
>>
>> (The plugin is available in JIRA in 
>> TS-1596<https://issues.apache.org/jira/browse/TS-1596%E2%80%8E>and on
>> Github <https://github.com/wkl/channel_stats>.)
>>
>> My question is whether there's any way to solve this so that counters
>> would be updated regularly and not only when completing the file download.
>>
>>
>> Also, on another note, is there any known issues in the "*
>> read-while-write*" option?  It seems not to work. I'm downloading a
>> large file via a number of different requests and only after the first
>> request is completed, the others are served too.
>>
>> Thanks,
>> Ron
>>
>
>
>
> --
> Mark Harrison
> Lead Site Reliability Engineer
> OmniTI
>

Reply via email to