Managing Application Resource Limits

Drycc Workflow supports restricting memory and CPU shares of each process. Requests/Limits set on a per-process type are given to Kubernetes as a requests and limits. Which means you guarantee <requests> amount of resource for a process as well as limit the process from using more than <limits>. By default, Kubernetes will set <requests> equal to <limit> if we don't explicitly set <requests> value. Please keep in mind that 0 <= requests <= limits.

Limiting Memory

If you set a requests/limits that is out of range for your cluster, Kubernetes will be unable to schedule your application processes into the cluster!

Available units for memory are:

Unit Amount
B Bytes
K KiB (Power of 2)
M MiB (Power of 2)
G GiB (Power of 2)

Important

The minimum memory limit allowed is 4MiB.

Use drycc limits:set <type>=<value> to restrict memory by process type, where value can be <limit> or <request>/<limit> format :

$ drycc limits:set web=64M
Applying limits... done

=== indoor-whitecap Limits

--- Memory
web     64M

--- CPU
Unlimited

$ drycc limits:set cmd=32M/64M
Applying limits... done

=== outdoor-whitecap Limits

--- Memory
cmd     32M/64M

--- CPU
Unlimited

If you would like to remove any configured memory limits use drycc limits:unset web:

$ drycc limits:unset web
Applying limits... done

=== indoor-whitecap Limits

--- Memory
Unlimited

--- CPU
Unlimited

Limiting CPU

You can also use drycc limits:set <type>=<value> --cpu to restrict CPU shares, where value can be <limit> or <request>/<limit> format. CPU shares are tracked in milli-cores. One CPU core is equivalent to 1000 milli-cores. To dedicate half a core to your process, you would need 500 milli-cores or 500m.

Unit Amount
1000m 1000 milli-cores == 100% CPU core
500m 500 milli-cores == 50% CPU core
250m 250 milli-cores == 25% CPU core
100m 100 milli-cores == 10% CPU core
$ drycc limits:set web=250m --cpu
Applying limits... done

=== indoor-whitecap Limits

--- Memory
web     64M

--- CPU
web     250m

$ drycc limits:set web=1500m/2000m --cpu
Applying limits... done

=== indoor-whitecap Limits

--- Memory
web     64M

--- CPU
web     1500m/2000m

You can verify the CPU and memory limits by inspecting the application process Pod with kubectl:

$ drycc ps
=== indoor-whitecap Processes
--- web:
indoor-whitecap-v14-web-8slcj up (v14)
$ kubectl --namespace=indoor-whitecap describe po indoor-whitecap-v14-web-8slcj
Name:       indoor-whitecap-v14-web-8slcj
Containers:
    QoS Tier:
      cpu:     Guaranteed
      memory:  Guaranteed
    Limits:
      cpu:     2000m
      memory:  64Mi
    Requests:
      memory:  64Mi
      cpu:     1500m

Important

If you restrict resources to the point where containers do not start, the limits:set command will hang. If this happens, use CTRL-C to break out of limits:set and use limits:unset to revert.

To unset a CPU limit use drycc limits:unset web --cpu:

$ drycc limits:unset web --cpu
Applying limits... done

=== indoor-whitecap Limits

--- Memory
Unlimited

--- CPU
Unlimited