SNMP

[Colton Conor]

What is the easiest was to use Prometheus with SNMP devices? We use over 20 different network vendors like Juniper, Cisco, Nokia, etc, and multiple models/devices under each brand. Each one of these has custom MIBs, and we are not SNMP experts. So we need a easy way to add multiple SNMP devices. 

[Brian Candler]

You can use file_sd_config:

- labels:

    module: mib_foo

  targets:

    - 1.2.3.4

    - 1.2.3.5

- labels:

    module: mib_bar

  targets:

    - 1.2.3.6

    - 1.2.3.7

Then use a relabelling config to copy label "module" to "__param_module"

  - job_name: 'snmp'

    scrape_interval: 15s

    file_sd_configs:

      - files:

        - /etc/prometheus/targets.d/snmp_targets.yml

    metrics_path: /snmp

    relabel_configs:

      - source_labels: [__address__]

        target_label: instance

      - source_labels: [__address__]

        target_label: __param_target

      - source_labels: [module]

        target_label: __param_module

      - target_label: __address__

        replacement: 127.0.0.1:9116  # SNMP exporter

This will let you define a single prometheus job which can query many different types of device.

It's still going to be your job to use the generator tool to convert MIBs into snmp.yml form suitable for use with snmp_exporter.  Sorry, there's no avoiding that.

[Ben Kochie]

90% of what you need from all the vendors is `if_mib`. The basic network traffic stats.

Once you have that, you can start to learn more about the vendor-specific MIBs.

On Sun, May 3, 2020 at 4:40 PM Colton Conor <colton...@gmail.com> wrote:

What is the easiest was to use Prometheus with SNMP devices? We use over 20 different network vendors like Juniper, Cisco, Nokia, etc, and multiple models/devices under each brand. Each one of these has custom MIBs, and we are not SNMP experts. So we need a easy way to add multiple SNMP devices. 

--
You received this message because you are subscribed to the Google Groups "Prometheus Users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to prometheus-use...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/prometheus-users/d40e0fa4-d44f-4bda-a753-5bea821b05f6%40googlegroups.com.

[Colton Conor]

I am just wondering how hard to would be to replace something like LibreNMS or PRTG with Prometheus? With Libre, we just install, set the SNMP community sting, scan our network, and all devices get mapped, linked, and monitored. I imagine Prometheus would be much more involved? Sounds like we would have to edit .yml files for each device / vendor just to get the SNMP metrics into the database. Then, we must create custom Grafana dashboards for each device type right?

[Ben Kochie]

Yes, you'll have to create the customizations yourself. There's currently no scanning/auto-configure for Prometheus SNMP users out there. It's a missing market opportunity.

I even suggested this to the LibreNMS people, as they have that part down solid, but their metrics collection and backend is at least a decade out of style (PHP polling + RRA).

The up-side is that Prometheus is easily more scaleable than LibreNMS. When I talked to them originally, I think we were still in the Prometheus 1.x days and estimated we were doing 5x better back then. Now it's probably 50x better.

[Colton Conor]

Ben,

Thanks for the information. I was really hoping you were going to say check out project x that has this all done for you :(. 

It sounds like today Promethus would just be too much work for our use case. With that being said, I see some NMS platforms like LibreNMS has a promethus exporter, which they claim is alpha at best. What are the thoughts about using something like this as a stepping stone to eventually getting to Promethus? 

Which monitoring platforms, other than LibreNMS, do you like or recommend? Do any of them have advanced functionality related to Promethus or time series databases? 

[Ben Kochie]

I don't actually recommend LIbreNMS. It's too slow and is too outdated in terms of monitoring capability to be worth implementing. Yes, it's easier to setup for _just_ SNMP monitoring. But it fails at everything else. Scalability, flexibility, performance, etc.

Prometheus is a universal platform for monitoring everything from network to server to application. Yes, it's more up-front work. But it's flexible enough to provide monitoring for everything in your organization.

To view this discussion on the web visit https://groups.google.com/d/msgid/prometheus-users/CAMDdSzPnyEZn2YtGtvkCcg6-mUdyiw%2Bi6gg%3DEAdMKev5SKxC8w%40mail.gmail.com.

[Colton Conor]

Ben,

I agree with you on  LIbreNMS being too slow and is too outdated. Do you have any recommendations other than that, or do you feel that all systems are too slow and is too outdated compared to Prometheus? In other words, who do you think are Prometheus's primary competitors? 

[Bruce]

you can try influxdb 

Le lundi 4 mai 2020 20:25:14 UTC+1, Colton Conor a écrit :

Ben,

I agree with you on  LIbreNMS being too slow and is too outdated. Do you have any recommendations other than that, or do you feel that all systems are too slow and is too outdated compared to Prometheus? In other words, who do you think are Prometheus's primary competitors? 

On Mon, May 4, 2020 at 1:37 PM Ben Kochie <sup...@gmail.com> wrote:

I don't actually recommend LIbreNMS. It's too slow and is too outdated in terms of monitoring capability to be worth implementing. Yes, it's easier to setup for _just_ SNMP monitoring. But it fails at everything else. Scalability, flexibility, performance, etc.

Prometheus is a universal platform for monitoring everything from network to server to application. Yes, it's more up-front work. But it's flexible enough to provide monitoring for everything in your organization.

On Mon, May 4, 2020 at 8:27 PM Colton Conor <colto...@gmail.com> wrote:

Ben,

Thanks for the information. I was really hoping you were going to say check out project x that has this all done for you :(. 

It sounds like today Promethus would just be too much work for our use case. With that being said, I see some NMS platforms like LibreNMS has a promethus exporter, which they claim is alpha at best. What are the thoughts about using something like this as a stepping stone to eventually getting to Promethus? 

Which monitoring platforms, other than LibreNMS, do you like or recommend? Do any of them have advanced functionality related to Promethus or time series databases? 

On Sun, May 3, 2020 at 4:43 PM Ben Kochie <sup...@gmail.com> wrote:

Yes, you'll have to create the customizations yourself. There's currently no scanning/auto-configure for Prometheus SNMP users out there. It's a missing market opportunity.

I even suggested this to the LibreNMS people, as they have that part down solid, but their metrics collection and backend is at least a decade out of style (PHP polling + RRA).

The up-side is that Prometheus is easily more scaleable than LibreNMS. When I talked to them originally, I think we were still in the Prometheus 1.x days and estimated we were doing 5x better back then. Now it's probably 50x better.

On Sun, May 3, 2020 at 11:10 PM Colton Conor <colto...@gmail.com> wrote:

I am just wondering how hard to would be to replace something like LibreNMS or PRTG with Prometheus? With Libre, we just install, set the SNMP community sting, scan our network, and all devices get mapped, linked, and monitored. I imagine Prometheus would be much more involved? Sounds like we would have to edit .yml files for each device / vendor just to get the SNMP metrics into the database. Then, we must create custom Grafana dashboards for each device type right?

On Sun, May 3, 2020 at 1:25 PM Ben Kochie <sup...@gmail.com> wrote:

90% of what you need from all the vendors is `if_mib`. The basic network traffic stats.

Once you have that, you can start to learn more about the vendor-specific MIBs.

On Sun, May 3, 2020 at 4:40 PM Colton Conor <colto...@gmail.com> wrote:

What is the easiest was to use Prometheus with SNMP devices? We use over 20 different network vendors like Juniper, Cisco, Nokia, etc, and multiple models/devices under each brand. Each one of these has custom MIBs, and we are not SNMP experts. So we need a easy way to add multiple SNMP devices. 

--
You received this message because you are subscribed to the Google Groups "Prometheus Users" group.

To unsubscribe from this group and stop receiving emails from it, send an email to promethe...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/prometheus-users/d40e0fa4-d44f-4bda-a753-5bea821b05f6%40googlegroups.com.

--
You received this message because you are subscribed to the Google Groups "Prometheus Users" group.

To unsubscribe from this group and stop receiving emails from it, send an email to promethe...@googlegroups.com.