Jobs in Foxx Queue are always "pending" in both Cluster and Single Server

[lanluu]

My Environment

Cluster

• ArangoDB Version: 3.4 RC4
• Deployment Mode: Cluster
• Infrastructure: Kubernetes; GKE; 4 Nodes; 1 vCPUs; 1.75 GB each Node and Kubernetes; GKE; 2 Nodes; 2 vCPUs; 7.5 GB each Node

Deployment YAML: I tried once with ---javascript.v8-contexts=4 and once without. Same results!

apiVersion: "database.arangodb.com/v1alpha"
kind: "ArangoDeployment"
metadata:
 name: "arangodb"
spec:
 mode: Cluster
 image: arangodb/arangodb-preview:v3.4.0-rc.4
 coordinators:
  count: 3
  args:
   - --javascript.v8-contexts=4

Single Server

• ArangoDB Version: 3.4 RC3 and 3.4 RC4
• Deployment Mode: Single Server
• Infrastructure: both Kubernetes; GKE; 1 Node; 2 vCPUs; 7.5 GB and Windows 10

Deployment YAML:

apiVersion: "database.arangodb.com/v1alpha"
kind: "ArangoDeployment"
metadata:
 name: "arangodb"
spec:
 mode: Single
 image: arangodb/arangodb-preview:v3.4.0-rc.4

Issue

Jobs in Foxx Queue are always pending i.e they are not executed.
Queue has 1 Worker.
I already setup another stronger Cluster with more CPUs, but it did not help.

[pluma]

Is the service responsive to requests otherwise? Can you maybe send us a zip bundle of your service for further testing?

[pluma]

Another possibility is that Foxx queues are simply disabled, but that seems unlikely.

[neunhoef]

I have checked that the foxx queues are activated in k8s, actually, I can even run queues. The job is executed on exactly one coordinator. I think we need more concrete information to reproduce this.

[lanluu]

@pluma Unfortunately I am not allowed to send you the Foxx Zip file as it contains license-based files. But actually, I tested by adding a very very simple Route to enqueue a Job which should be triggered immediately. The job does nothing, except calls console.log. Right now, I cannot reproduce such of problem in Single Server with '3.4-RC4', but I can see this problem on Cluster (and 3.4-RC4). Sometimes the Job is executed and sometimes it is ... not. There was two times, in Single Server, Job was marked as in progress but it did nothing :-(.

[pluma]

@lanluu I've created a service that simply logs a message to the console when executed, which @neunhoef has used with the results above. Can you give it a try and see if you observe the same behavior?

queue-test.zip

[maxkernbach]

Hi @lanluu,

Did you already have a chance to test the provided service queue-test.zip?

[maxkernbach]

Since we have not heard back from you in a while, I am closing this issue as "cannot reproduce" for now.

In case the problem persists, please feel free to add a comment and we will reopen this issue.