🐛 [Bug] The Network Gives A 504 Gate-Away Time-Out From Time To Time

[Tirokk]

🐛 Bugreport

Sometimes the network gives a 504 gate-way time-out.
(Heavily happened on the 13th and 14th of June 2020 and in the week before as well.)

There was nor explicit error I’ve seen. Except an event I got displayed by:
kubectl -n human-connection get events --sort-by=.metadata.creationTimestamp
It said:
Warning – FailedScheduling – pod/nitro-neo4j-86596f49cd-snjn4 – 0/7 nodes are available: 6 node(s) didn't match node selector, 7 Insufficient memory.

At the end I had to restart everything several time by deleting all the network pods on after the other.

Matt and I solved a memory problem of the Neo4j database by applying memory requests and limits to the containers of the pods. We labeled a new node with enough memory and assigned the Neo4j pod to the label of this node, as to say to the node.
See issue #2628 and PR #2629 .

I have the impression that this could help us to solve the network problem, because.
It is recommended anyways:

• https://kubernetes.io/docs/concepts/configuration/manage-compute-resources-container/
• https://cloud.google.com/blog/products/gcp/kubernetes-best-practices-resource-requests-and-limits

We have to define the following settings to all deployments (Webapp, Backend, Neo4j):

    spec:
      containers:
        resources:
          requests:
            memory: XXX
            cpu: XXX
          limits:
            memory: XXX
            cpu: XXX

Steps to reproduce the behavior

Happens from time to time …

Expected behavior

Network should not give an error …

Additional context

None.