Acumos Boreas docker-proxy and docker-dind fails when using the installation procedures

Question

I have tried to setup Acumos Boreas on a single Ubuntu 18.04 Virtual Machine (with 12 CPU's, 32 GB RAM and 300 GB Disk space) using the following procedure:

0) git clone https://gerrit.acumos.org/r/system-integration

1) cd system-integration/tools

2) bash setup_docker.sh

Add user to docker group: sudo usermod -aG docker $USER and logout/in again

3) bash system-integration/AIO/acumos_k8s_prep.sh $USER acumos.tele.no 2>&1 | tee aio_prep.log

4) cd system-integration/AIO

5) bash oneclick_deploy.sh 2>&1 | tee oneclick_deploy.log

The installation fails.

Looking at the output from the kubectl commands (and info in the Kubernetes Dashboard) I have observed the following:

1) I see a lot of this: Pod The node was low on resource: [DiskPressure] , which is strange since the node has plenty free diskspace and inodes.

2) The docker-proxy and docker-dind service fails to start the associated pod multiple times, the docker-proxy finally "succeeds" to run but errors in logfile (see below).

3) The docker-dind-xxxx pod fails completely to start up: State: Waiting, Reason: CrashLoopBackOff, Last State: Terminated, Reason: Error

In the logfile for the docker proxy I see this:

Reading state information...
E: Unable to locate package python
E: Unable to locate package python-pip
Waiting for temporary DNS resolution issue to be fixed
Reading package lists...
Building dependency tree...
Reading state information...
E: Unable to locate package python
E: Unable to locate package python-pip
Waiting for temporary DNS resolution issue to be fixed
Reading package lists...
Building dependency tree...
Reading state information...
E: Unable to locate package python
E: Unable to locate package python-pip
Waiting for temporary DNS resolution issue to be fixed
Reading package lists...
Building dependency tree...
Reading state information...
E: Unable to locate package python
E: Unable to locate package python-pip

Any ideas on how to solve this?

Answer 1

The docker-dind issue is one I just found as well. The upstream project has changed how they expose docker services. I am working on a patch right now to fix it. But you can edit your docker-proxy-deployment.yaml (in AIO/docker-engine/kubernetes), to reference the image version docker:18-dind for both containers, ie "image: docker:18-dind". That fixes it (previously it was referencing the latest docker:dind image, which caused the issue). After you edit the template, re-run the AIO/docker-engine/setup_docker_engine.sh script.

The docker-proxy issue is less clear to me. Typically I see such DNS resolution issues as an inherent thing you will encounter, sporadically, in k8s envs. There are well-documented issues with DNS reliability in k8s, related to UDP request collisions. For now we get by through retrying where needed. But usually the issue resolves itself soon.