topic Query Kuberentes workloads in DQL

Query Kuberentes workloads

DamianJankowski — Fri, 28 Mar 2025 12:52:24 GMT

Hi,

Could you please help me better understand how to query entities and the relationships between them?

fetch dt.entity.service
| filter in(id, classicEntitySelector("type(service),fromRelationship.isServiceOf(type(CLOUD_APPLICATION),entityId(\"CLOUD_APPLICATION-D40C6BABEC087314\"))"))

This returns all services related to a specific cloud application. In my case, it includes:

-Requests executed in background threads of SpringBoot for dine-in-api-*

-Web-server for dine-in-api-*

What I want to achieve is to get the Kubernetes workload names, which are visible in the “services” application (either in the filter bar or at the top of the service page). I plan to use these names as variables in a dashboard to display metrics like response time or failure rate, split by each service.

Re: Query Kuberentes workloads

krzysztof_hoja — Sun, 29 Sep 2024 19:54:48 GMT

You query is correct. We just need to make it "dynamic" and use filtering criteria "by name" instead of "by id".
In combination with getting metric you mention it will look like:

timeseries response_time=avg(dt.service.request.response_time), by: {dt.entity.service}, filter: in( dt.entity.service, classicEntitySelector(concat("type(service),fromRelationship.isServiceOf(type(CLOUD_APPLICATION),entityName(",$Entity,"))"))) | fieldsAdd dt.entity.service.name = entityName(dt.entity.service)

As dimensions for this metric contains id of the service, you can add service name using entityName function

Re: Query Kuberentes workloads

DamianJankowski — Mon, 30 Sep 2024 09:49:42 GMT

Hi @krzysztof_hoja

thanks a lot.

I would say it is only a part of the desired solution, so first I would like to get the pure Kubernetes workload name:

Kubernetes workload name is going to be a variable,

then I would like to query based on that name (e.g. to get 5xx timeseries) to get all related services.

btw. is it possible to trim service-id from your query as I am getting sth like: SERVICE-86915F82C40F74B8•payments-api PaymentLinksController

Re: Query Kuberentes workloads

krzysztof_hoja — Mon, 30 Sep 2024 10:08:56 GMT

To get list of workload names to use in variable definition:

fetch dt.entity.cloud_application | fields entity.name | dedup entity.name

Of course you can control what fields are in final results using fields of fieldsRemove commands.

timeseries response_time=avg(dt.service.request.response_time), by: {dt.entity.service}, filter: in( dt.entity.service, classicEntitySelector(concat("type(service),fromRelationship.isServiceOf(type(CLOUD_APPLICATION),entityName(",$Entity,"))"))) | fields dt.entity.service = entityName(dt.entity.service), response_time, interval, timeframe

Re: Query Kuberentes workloads

DamianJankowski — Mon, 30 Sep 2024 12:52:08 GMT

@krzysztof_hoja thanks, second query indeed is solving the issue, thanks a lot.

is there an other option to list workloads?

Re: Query Kuberentes workloads

krzysztof_hoja — Mon, 30 Sep 2024 16:50:29 GMT

If you want data related variable, one fed by DQL results is the easiest one to use.

Re: Query Kuberentes workloads

DamianJankowski — Mon, 30 Sep 2024 20:56:37 GMT

Hmm… the workload name (without a lot of trash) should be usable as a variable to create a tile that queries all related services

Re: Query Kuberentes workloads

krzysztof_hoja — Tue, 01 Oct 2024 10:25:35 GMT

I am not sure what is your question here....

Query retrieving e.g. response time takes workload name ($Entity matching name of CLOUD_APPLICATION).

Re: Query Kuberentes workloads

DamianJankowski — Tue, 01 Oct 2024 16:52:15 GMT

@krzysztof_hoja I would like a query that returns only the Kubernetes workloads and another one that returns all entities belonging to a specific workload. So the question is - how to achieve that?

Re: Query Kuberentes workloads

AurelienGravier — Wed, 26 Mar 2025 19:09:22 GMT

Hi @krzysztof_hoja , thanks for sharing!

Do you know if there's an alternative approach to correlate metrics like service response time with total CPU consumption of the pods belonging to its associated workloads?

It works with classicEntitySelector but I can't iterate on each entity without coding or a workflow :

fetch dt.entity.cloud_application | filter in( id, classicEntitySelector("type(CLOUD_APPLICATION),toRelationships.isServiceOf(type(SERVICE),entityId(SERVICE-B82BE29035866C21))")) | fieldsAdd serviceID = "SERVICE-B82BE29035866C21"

I'm looking for a way to build an iterative view ideally suitable for rendering in a notebook table, where each row would contain:
- Service name
- Avg response time
- Total CPU usage of its workload’s pods (aggregated)

Would love to know if you've explored something similar, or have any tips on composing the entity graph to support this use case.

Thanks you very much.
Regard Aurélien.

Re: Query Kuberentes workloads

AurelienGravier — Fri, 28 Mar 2025 06:16:07 GMT

Hello @krzysztof_hoja ,

I eventually figured out how to do this iteratively using the belongs_to method :

fetch dt.entity.service | filter isnotnull(clustered_by) // Get namespaces from service | fieldsAdd namespaceId = belongs_to[dt.entity.cloud_application_namespace] | expand namespaceId | fieldsAdd namespace = entityAttr(namespaceId, "entity.name", type:"dt.entity.cloud_application_namespace") // Get workloads from service | fieldsAdd workloadId = belongs_to[dt.entity.cloud_application] | expand workloadId | fieldsAdd workloadName = entityAttr(workloadId, "entity.name", type:"dt.entity.cloud_application") // get service median response time | lookup [ timeseries responseTime_p50_timeseries = percentile(dt.service.request.response_time, 50), by:{dt.entity.service}, bins: 40 ], sourceField:id, lookupField:dt.entity.service, fields:{responseTime_p50_timeseries}, executionOrder:leftFirst | fieldsAdd responseTime = arrayLast(responseTime_p50_timeseries) // Add CPU % Usage per workload // Only work for workload with a limit set | lookup [ timeseries by:{dt.entity.cloud_application}, { cpu_usage = sum(dt.kubernetes.container.cpu_usage, default:0, rollup:avg), cpu_limit = sum(dt.kubernetes.container.limits_cpu, rollup:avg) } | fieldsAdd cpuPercent = (arrayAvg(cpu_usage) / arrayAvg(cpu_limit)) * 100 ] , sourceField:workloadId, lookupField:dt.entity.cloud_application, fields:{cpuPercent} | fieldsAdd cpuPercent = if(isNull(cpuPercent), "no limit", else: cpuPercent) ....

Thank you !