[function-mesh] nil pointer in MakeFunctionHPA()

[imaffe]

A nil pointer deference is thrown in

The exception is thrown because function.Spec.MinReplica is null:

2022-11-02T15:54:14.680+0800 INFO function-resource MinReplicas {"pt": null}

The spec json is

{
   "specs":{
      "className":"org.apache.pulsar.functions.api.examples.ExclamationFunction",
      "clusterName":"test-pulsar",
      "replicas":1,
      "maxReplicas":5,
      "input":{
         "typeClassName":"java.lang.String",
         "topics":[
            "persistent://public/default/java-function-input-topic"
         ]
      },
      "output":{
         "typeClassName":"java.lang.String",
         "topic":"persistent://public/default/java-function-output-topic"
      },
      "logTopic":"persistent://public/default/logging-function-logs",
      "resources":{
         "limits":{
            "cpu":"200m",
            "memory":"1100M"
         },
         "requests":{
            "cpu":"100m",
            "memory":"1G"
         }
      },
      "secretsMap":{
         "name":{
            "path":"test-secret",
            "key":"username"
         },
         "pwd":{
            "path":"test-secret",
            "key":"password"
         }
      },
      "volumeMounts":[
         {
            "name":"cache-volume",
            "mountPath":"/cache"
         }
      ],
      "autoAck":true,
      "forwardSourceMessageProperty":true,
      "maxPendingAsyncRequests":1000,
      "pod":{
         "labels":{
            "locaction":"mtv"
         },
         "annotations":{
            "managed-function":"true"
         },
         "volumes":[
            {
               "name":"cache-volume",
               "emptyDir":{
                  
               }
            }
         ],
         "imagePullSecrets":[
            {
               "name":"regcred"
            }
         ],
         "initContainers":[
            {
               "name":"init-function",
               "image":"busybox:1.28",
               "command":[
                  "sh",
                  "-c",
                  "echo The app is running! && sleep 30"
               ],
               "resources":{
                  
               }
            }
         ],
         "sidecars":[
            {
               "name":"sidecar-function",
               "image":"busybox:1.28",
               "command":[
                  "sh",
                  "-c",
                  "echo The app is running! && sleep 30000"
               ],
               "resources":{
                  
               }
            }
         ],
         "env":[
            {
               "name":"EXAMPLE_VARIABLE",
               "value":"exampleValue"
            }
         ]
      },
      "pulsar":{
         "pulsarConfig":"test-pulsar",
         "authSecret":"test-auth",
         "tlsSecret":"test-tls"
      },
      "java":{
         "jar":"pulsar-functions-api-examples.jar",
         "jarLocation":"public/default/nlu-test-java-function",
         "extraDependenciesDir":"random-dir/"
      }
   }
}

I can follow this issue to understand the HPA related code~

[imaffe]

Looks like my default webhook is not enabled.

[tpiperatgod]

Currently, this issue occurs during upgrades from older versions (without webhook) to newer versions. Therefore, it makes sense to solve this issue.

[imaffe]

This is because, when I run the operator controller locally using make run,seems the webhook is not created and thus the nil pointer. (If webhook enabled, the validation should guarantee that the nil pointer condition will not occur). Need to find a way to enable webhook in local run.

[imaffe]

Currently, this issue occurs during upgrades from older versions (without webhook) to newer versions. Therefore, it makes sense to solve this issue.

Good to know this. For fixing this issue, do you mean something like check if the pointer is nil and provide a default value for it before we process the Spec (just like we did in the mutating webhook)?

Or should we just simply detect the nil pointer and fail early and instruct the user to complete the Spec ?

[tpiperatgod]

This is because, when I run the operator controller locally using make run,seems the webhook is not created and thus the nil pointer. (If webhook enabled, the validation should guarantee that the nil pointer condition will not occur). Need to find a way to enable webhook in local run.

It is recommended to use nocalhost to run the controller with webhook enabled, pls see: https://streamnative.slack.com/files/U03EV808CAC/F03RZTPPWE8/develop_functionmesh_with_nocalhost.mp4

[tpiperatgod]

Good to know this. For fixing this issue, do you mean something like check if the pointer is nil and provide a default value for it before we process the Spec (just like we did in the mutating webhook)?

Or should we just simply detect the nil pointer and fail early and instruct the user to complete the Spec ?

I think the "default value" is better...

[imaffe]

Thanks, I'll try with the nocalhost~