So, an update here:

Gathering CPU + 3 SNMP OIDs:

23:20:30.342653 IP 10.0.X.X.59192 > 10.0.Z.Z.25826: UDP, length 1308
23:21:20.341185 IP 10.0.X.X.59192 > 10.0.Z.Z.25826: UDP, length 1308
23:22:10.341337 IP 10.0.X.X.59192 > 10.0.Z.Z.25826: UDP, length 1308
23:23:00.341386 IP 10.0.X.X.59192 > 10.0.Z.Z.25826: UDP, length 1308

(50s approx)

Now, I added the memory. (CPU + 3 SNMP OIDs + MEMORY)

New results:

23:24:11.226821 IP 10.0.X.X.47903 > 10.0.Z.Z.25826: UDP, length 1309
23:24:31.288530 IP 10.0.X.X.47903 > 10.0.Z.Z.25826: UDP, length 1293
23:25:01.273344 IP 10.0.X.X.47903 > 10.0.Z.Z.25826: UDP, length 1318
23:25:31.257782 IP 10.0.X.X.47903 > 10.0.Z.Z.25826: UDP, length 1300

(30s approx)

Now, I added interface and disk. (CPU + 3 SNMP OIDs + MEMORY + DISK +
INTERFACE)

23:26:31.466675 IP 10.0.X.X.48354 > 10.0.Z.Z.25826: UDP, length 1328
23:26:41.465791 IP 10.0.X.X.48354 > 10.0.Z.Z.25826: UDP, length 1301
23:26:51.465875 IP 10.0.X.X.48354 > 10.0.Z.Z.25826: UDP, length 1306
23:27:01.511527 IP 10.0.X.X.48354 > 10.0.Z.Z.25826: UDP, length 1292

(10s approx)

This obviously means the more data, the faster you fill the buffer, the
more often you send the metrics.

All these tests are with the default value for queues/buffers/limits.

I am going to play with those values now and compare the results to these.


I'll update again later.

Joel.

On Tue, Feb 16, 2016 at 3:04 PM, Giovanni Torres <giovanni.tor...@gmail.com>
wrote:

>
>
> On Tue, Feb 16, 2016 at 1:22 PM Joel Serrano <j...@gogii.net> wrote:
>
>> Well I really think it could be what Marc said.
>>
>> On v5.5 until the buffer is full it won't wend anything (anywhere...).
>>
>> I'm going to lower the buffer and see if the reporting time decreases, if
>> Marc is right, it should work.
>>
>> I'll post back the results.
>>
>
> I'm curious to hear your results.  How much did you lower the buffer size
> to vs how many metrics are you handing?
>
> Giovanni
>
_______________________________________________
collectd mailing list
collectd@verplant.org
http://mailman.verplant.org/listinfo/collectd

Reply via email to