Service request counts per host server

[CharlesPerez]

Hi All,

I would like to know if it is possible to split the service request counts of a multi-hosted service, per host server.

For example: AppServer 1, AppServer 2, AppServer 3, AppServer4, with Service1 running on all four servers
I want to create a table that will look like this

Host	Service	Request Count
AppServer 1	Service1	56k
AppServer 2	Service1	54k
AppServer 3	Service1	81k
AppServer 4	Service1	59k

Do you know a DQL to get this?

[AntonPineiro]

Hi,

You can see that information already seeing requests:

You can see information per instance.

Best regards

[waikeat_chan]

I think perhaps Charles is asking  is it possible to do it via DQL

[CharlesPerez]

That is correct sir. Apologies, I was not able to mention it on my earlier inquiry.

Thank you for pointing that out 😊

Regards,
Charles

[AntonPineiro]

Hi,

I have just added DQL reference in message, and move it to DQL forum.

Best regards

[CharlesPerez]

Thank you for the assistance Sir @AntonPineiro!! 😊

[CharlesPerez]

Hi AntonPineiro,

Yes, we already saw this. But client requires to see the distribution of request count per host machine.

I can do this with processes, but client wants to do it with Services.

Regards,
Charles 

[AntonPineiro]

Hi,

I think @sia_h has already pointing you to a solution.

Best regards

[lubrman]

Hi @CharlesPerez 

You should be able to create a tag within each server’s process. Based on that tag, it should then be possible to divide individual services. This way, you can also create the following table.

[sia_h]

Try this:

fetch dt.entity.process_group_instance
| fieldsAdd service.id = runs [dt.entity.service], process.name = entity.name, belongs_to
| fieldsAdd dt.entity.host = belongs_to[dt.entity.host]
| fieldsAdd dt.entity.host.name = entityAttr(dt.entity.host, "entity.name")
| expand service.id
| lookup [
timeseries sum(dt.service.request.count),
by:{
dt.entity.service, 
dt.entity.process_group 
}
], sourceField:service.id, lookupField:dt.entity.service, prefix: "metric."

[CharlesPerez]

Hi @sia_h 

Thank you for this! Will simulate this first on my end and see if it matches our requirements.

Hopefully this is what they are looking for 😁. 

Regards,
Charles Perez 

[CharlesPerez]

Hi @sia_h,

I tried to simplify the table, but the service seems to appear Null.

Only the host name and process name have data.  Would like to seek help on how I can fix this. Thank you! 

Regards,
Charles Perez

[sia_h]

You could add 

| filter isNotNull(service.id)

to the end of the query. Processes with a null Service Name either have no services, or had no services in the timeframe you're looking in.

[CharlesPerez · Accepted solution]

Hi Everyone,

I have resolved the Query. Thank you @AntonPineiro , @sia_h , @waikeat_chan , @lubrman  for the assistance.

Pasting here the final DQL that I have used, and the result from the Demo Environment.

Hope this helps other engineers! 😁

DQL:

timeseries by:{dt.entity.service}, request_count = sum(dt.service.request.count)
| summarize by:{dt.entity.service}, `Request Count` = sum(arraySum(request_count))
| fieldsAdd `Service Name` = entityAttr(dt.entity.service, "entity.name")
| lookUp
[
fetch dt.entity.service
| expand dt.entity.host=runs_on[dt.entity.host]
| fieldsAdd dt.entity.host.name = entityAttr(dt.entity.host, "entity.name")
], lookupField:id, sourceField:dt.entity.service
| fieldsRename `Host Name` = lookup.dt.entity.host.name
| filter isNotNull(`Host Name`)
| fields `Host Name`, `Service Name`, `Request Count`

Result:

Regards,
Charles Perez

[imrj99]

looking for something similar but need to aggregate or sum up all the requests count for each host (disregarding service name, not needed), so that the table would just have a single count value per server