Msg : Metrics updated

[aparnareji]

Hi @kbudde ,

I have successfully deployed a container in my environmnet for rabbitmq-exporter . The pod is also running , but am getting the following message :

time="2018-03-06T06:56:28-05:00" level=info msg="Metrics updated" duration=4.424415629s

And also am not able to view the metrics , as am getting the error :

504 Gateway Time-out

Hope this is similar to the issue #6 .
Please note that , I am using the latest version and exposed port 9419

Could you please help me to solve this.

With Regards,
Aparna Reji

[chetan-hireholi]

Hello @aparnareji ,
We interacted in the CloudFoundry exporter issue as well. I think, even I'm trying to setup a similar thing.
Can you please let me know how are you running the exporter?

[aparnareji]

Hi @SamwiseGambgee ,

Am actually deploying the exporter in Openshift , inside my project and then will target the metrics to Prometheus.

With Regards,
Aparna Reji

[chetan-hireholi]

@aparnareji oh i see...
In my case, I need to tunnel(ssh) it in order to connect to AWS.
Do you have any idea how can I achieve it?

[aparnareji]

@SamwiseGambgee ,

As of now , am not using AWS. Sorry, I have no idea about it .

With Regards,
Aparna Reji

[aparnareji]

Hi Team,

Could you please give me an update .

With Regards,
Aparna Reji

[kbudde]

Hi,

The first log message "metrics updated" is normally a good sign. On the other hand 4s for generating the metrics is a very long time.

Can you please give me some more details:

• Which configuration is in use for the exporter? (I recommend bert, no_sort for the features)?
• How many queues and exchanges do you have?
• Which rabbitmq version are you using (single node/cluster)?

updated:
Is it possible that the metrics endpoint of the xporter is not reachable in your setup?

[aparnareji]

Hi @kbudde ,
`

• How many queues and exchanges do you have?
• ---It is dynamic ( for now: Queue-7379, exchanges-293)
• Which rabbitmq version are you using (single node/cluster)
• ----Single node
• Which configuration is in use for the exporter? (I recommend bert, no_sort for the features)?
• ---I didnt get the question , i have now deployed the exporter in openshift
  `
  Now am getting detailed log with same message:

time="2018-03-14T07:38:12-04:00" level=error msg="Error while retrieving data from rabbitHost" error="Get <rabbiturl>/api/queues:  net/http: request canceled (Client.Timeout exceeded while awaiting headers)" host="<rabbit URL> " statusCode=0

and

time="2018-03-14T07:38:12-04:00" level=info msg="Metrics updated" duration=10.25768936s

The metrics endpoint of the exporter is reachable in our setup

With Regards,
Aparna Reji

[aparnareji]

Hi @kbudde ,

Could you please give me an update

With Regards,
Aparna Reji

[kbudde]

Hi,

Are you sure that the exporter can reach the rabbitmq?
You need as well the management plugin installed on rabbitmq.

[aparnareji]

Hi @kbudde ,

Am sure that the exporter can reach the rabbitmq as sometimes it connect properly and gives the metrics. But sometimes, i will get this error and then there wont be any metrics and I also have the management plugin installed on rabbitmq .

With Regards,
Aparna Reji

[kbudde]

HI @aparnareji,

what is the normal scrape duration for your setup if it working?

 level=info msg="Metrics updated" duration=???s

The default timeout is 10s. I think this could be the problem.
Are you using RABBIT_CAPABILITIES=bert,no_sort? This could reduce the scraping duration.

[aparnareji]

Hi @kbudde ,

sorry for the delayed reply.

The duration , now having is duration=10.131007724s
RABBIT_CAPABILITIES=bert,no_sort is not used
With Regards,
Aparna Reji

[kbudde]

Hi,

for the moment I've increased the timout for the client to 15s.
But I'm going to add a better logic (prometheus is providing timeout information during scraping) to the exporter later.
For your usecase the bert support and no_sort should significantly reduce cpu load during scaping and normally should speed up the process as well.
I really recommend using it.

[aparnareji]

Hi @kbudde ,

Thank you. May i please know what is the job of bert and no_sort.

With Regards,
Aparna Reji

[kbudde]

Hi,

Extended RabbitMQ capabilities

Newer version of RabbitMQ can provide some features that reduce
overhead imposed by scraping the data needed by this exporter. The
following capabilities are currently supported in
RABBIT_CAPABILITIES env var:

• no_sort: By default RabbitMQ management plugin sorts results using
  the default sort order of vhost/name. This sorting overhead can be
  avoided by passing empty sort argument (?sort=) to RabbitMQ
  starting from version 3.6.8. This option can be safely enabled on
  earlier 3.6.X versions, but it'll not give any performance
  improvements. And it's incompatible with 3.4.X and 3.5.X.
• bert: Since 3.6.9 (see
  Add support for non-JSON encoding of replies rabbitmq/rabbitmq-management#367) RabbitMQ
  supports BERT encoding as a JSON alternative. Given that BERT
  encoding is implemented in C inside the Erlang VM, it's way more
  effective than pure-Erlang JSON encoding. So this greatly reduces
  monitoring overhead when we have a lot of objects in RabbitMQ.

[aparnareji]

Hi @kbudde ,

Thanks for your update , may i give the env variable no_sortwith value ./rabbitmq_exporter

With Regards,
Aparna Reji

[kbudde]

HI @aparnareji ,

at the moment just environment variables are supported (no flags).

Only solution is:

PUBLISH_PORT=9419 RABBIT_CAPABILITIES=bert,no_sort ./rabbitmq_exporter