Gauge-type data for default dashboard metrics

[maxlemieux]

Is your feature request related to a problem? Please describe.

The metrics for the default dashboard for the Varnish integration are of type 'rate'. This means when the Varnish server restarts, the metrics spike down and skew the graphs making the dashboard useless.

Feature Description

It would be great if we could also have 'gauge' metrics for the metrics which are on the dashboard, in order to have the dashboard by default behave as expected for server restarts.

This would affect at least the following metrics based on the default dashboard:

• session.connections
• net.requests
• sessions.drops
• main.expired
• lru.nuked
• cache.hits
• cache.misses
• cache.pass
• cache.grace
• cache.miss
• net.backend.requests
• backend.connections

Describe Alternatives

Default dashboard charts could exclude values under 0 as a workaround.

Additional context

This behavior is confusing for dashboard users.

Priority

Really Want

[matewilk]

Hey, sorry for chiming in, I'm not a maintainer of the repo, I've created a PR recently and just looking around and looked at your issue out of curiosity. Would writing an NRQL query that excludes 0s help or is it something more complex?

[maxlemieux]

Would writing an NRQL query that excludes 0s help or is it something more complex?

That is a good question and something worth considering - it could be a workaround for now, although it would require users to create new charts on the dashboard.

While this might help the dashboard charts preserve a useful Y-axis view - fixing the issue of skew from large negative rate values on service restarts - it would not seem optimal from a "truthiness" perspective, since if rate is what is being charted all rate metrics should probably be displayed.

If the default-created dashboards were to use a hypothetical new gauge metric for these attribute types instead, this could allow clearly showing all values without cleaning the data and potentially losing valuable information, in the sense that the large negative values expose the service restarts.

[carlossscastro]

https://github.com/newrelic/nri-varnish/pull/42

[carlossscastro]

@maxlemieux @matewilk We'll make a new release next monday to include this fix.

[maxlemieux]

Thanks Carlos! Will users need to update their existing dashboards to parse the new metric?

[carlossscastro]

@maxlemieux Users will need to update the integration to the latest version once it gets released. They do not need to update any dashboards, just the integration.

[kang-makes]

v2.5.1 released. It will take some time to have it in the infrastructure-bundle and Docker images but binaries/packages are ready :)