scale up with cpu utilization

[puspbhatt]

Is there a way that i could make this autoscalable with the cpu utilization.

[endeepak]

You can use any metric for autoscaling as long as that metric is capture in your metric store like prometheus.

Could you please provide more details on which metric you have captured in metric store and want to use for autoscaling

[puspbhatt]

i was trying to take cpu and memory uses of swarm but i couldn't find any metric_query to capture. for example when cpu and memory goes above 70% it should spin up a new container.

[endeepak]

Please take a look at https://github.com/google/cadvisor/blob/master/docs/storage/prometheus.md for capturing container metrics

…

On Wed, Oct 10, 2018, 4:47 AM puspbhatt ***@***.***> wrote: i was trying to take cpu and memory uses of swarm but i couldn't find any metric_query to capture. for example when cpu and memory goes above 70% it should spin up a new container. — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#6 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAS9ULJK5oRyIHK9rZ-vmJ3P0vdloSh4ks5ujS6RgaJpZM4XMLCp> .

[puspbhatt]

autoscale_rules:

• service_name: example_web
  scale_min: 1
  scale_max: 3
  scale_step: 1
  metric_store: monitoring
  metric_query: scalar(avg(rate(container_spec_cpu_period{job="web"}[30s])))
  scale_up_threshold: 70 %
  scale_down_threshold: 30%
  i changed these parameter in autoscaler.yml, but it didn't work

[endeepak]

Please check if "container_spec_cpu_period" is captured in prometheus and run this query in prometheus query console to check the values

…

On Wed, Oct 10, 2018 at 7:08 AM puspbhatt ***@***.***> wrote: autoscale_rules: - service_name: example_web scale_min: 1 scale_max: 3 scale_step: 1 metric_store: monitoring metric_query: scalar(avg(rate(container_spec_cpu_period{job="web"}[30s]))) scale_up_threshold: 70 % scale_down_threshold: 30% i changed these parameter in autoscaler.yml, but it didn't work — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#6 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAS9UBZ2T9mKIDRibUNLCwtdi4y_WJXwks5ujU9_gaJpZM4XMLCp> .

-- Regards, Deepak N @endeepak

[puspbhatt]

100 - (avg by (instance) (irate(node_cpu_seconds_total{job="node",mode="idle"}[5m])) * 100)
I found this query to capture the cpu utilization of the swarm, when i run this query in prometheus UI it says no data.
Can you please help me with that.

[puspbhatt]

what are the configuration changes that needs to be changed if we wanna autoscale for cpu uses. if possible can you please elaborate in bullet points, i am stuck when trying to set it up.
Thank you Deepak