Tried to run the BOINC container. Got an error message from docker

[Michael von Blucher af Altona]

Hi 

I tried to run the BOINC container using 

>docker run -d --name boinc --net=host -v /opt/appdata/boinc:/var/lib/boinc -e BOINC_GUI_RPC_PASSWORD="123" -e BOINC_CMD_OPTIONS="--allow_remote_gui_rpc" boinc/client

and Docker responded with error message:

>docker: Error response from daemon: unable to find "net_prio" in controller set: unknown.

Anyone have any ideas?

Regards

Michael Blucher.

[Michael von Blucher af Altona]

Hi

I've done a few searches on the web.  It seems this problem is well known and has 2 known solutions as follows:

• build a new kernel with CONFIG_CGROUP_NET_PRIO enable ?
• downgrade systemd ?
  

Both of the above options are beyond my knowledge.   Not even sure where I would start.

Does anyone know of any other solution?

Does anyone know when a fixed version of Docker will be available?

Has anyone out there already built a new Raspbian kernel that can be used to get around the problem?

Regards

Michael Blucher

[Michael von Blucher af Altona]

Hi All

No-one has replied to this problem, so I'm assuming there is no easy solution.

I figured I'd try upgrading my pi and pi zeros as per Chris Burton's upgrade post, i.e.

sudo svn --force export https://github.com/burtyb/clusterhat-image/trunk/files /

sudo apt-get update

sudo apt-get -y dist-upgrade

This worked.  The error below went away:

>docker: Error response from daemon: unable to find "net_prio" in controller set: unknown.

Except, now I am back at square one.

Docker version 19.03.3 installed on my pi controller and while it worked well on the controller,

Docker 19.03.3 did not install correctly on any of my pi zeros.  Docker installed with error and refused to run, i.e, daemon refused to start.

This is precisely where I was when I first installed Docker.

So I back rev'd Docker to version 18.06.1 which worked on the pi zeros, however Docker 19.03.3 did not like that.

It reported the Nodes correctly, but did not seem to want to actually work with them.

So I back rev'd Docker on the controller to version 18.06.1, which likes the pi zeros, but comes back with error:

>docker: Error response from daemon: unable to find "net_prio" in controller set: unknown.

Square one!

Does anyone know how I can successfully install Docker version 19.03.3 on my pi zeros?

That appears to be the hurdle.

Regards

Michael B.

On Sunday, September 29, 2019 at 8:41:33 PM UTC+10, Michael von Blucher af Altona wrote:

[Charles Taylor]

I haven't run docker, but I did find that your not alone with this error.
You might find it interesting as I found this on Docker forums

https://forums.docker.com/t/cannot-connect-to-the-docker-daemon-is-the-docker-daemon-running-on-this-host/8925

[Michael von Blucher af Altona]

Thanks Charles

That is howeer a different error, an error you get if you mess up your docker daemon port.  I have done that in the past and had to fix it.

The error I'm looking at here requires either a Kernell rebuild, back reving systemd or as I've just found out today, upgrading Docker to 19.03.1.

Cheers

Michael B

[Charles Taylor]

Yes did find that on other forums, I wish you all the best with your project.

[Michael von Blucher af Altona]

It seems Peter Cross has the same problem as I do with my pi zeros.

Neither of us could run Docker 19.03.1 on the pi zees

Question: Will the following allow me to run Docker 19.03.1 on my pi zees?

curl -s https://packagecloud.io/install/repositories/Hypriot/rpi/script.deb.sh | sudo bash

sudo apt-get install containerd.io=1.2.6-1

I already have Docker 18.06.1 on my pi zees.   Docker 18.06.1 has issues on my rpi controller.

What is the equivalent command (like below) but for installing Docker 19.03.1?

sudo apt-get install docker-ce=18.06.1~ce~3-0~raspbian

Once I have that, I can experiment on the weekend, stepping up my pi zees to 19.03.1.

If I can get it running on 1, I'll upgrade the whole cluster.

Cheers

Michael B.

[Peter Cross]

I got it to work last night with the below two commands. I just wanted to automate the iptables commands without having to manually do it every time. 

Cheers!

Peter Cross

Cell: 210-232-4748

Work: 720-443-8808

Sent from my iPhone

Please excuse any errors. 

--
You received this message because you are subscribed to the Google Groups "ClusterHAT" group.
To unsubscribe from this group and stop receiving emails from it, send an email to clusterhat+...@googlegroups.com.
To view this discussion on the web, visit https://groups.google.com/d/msgid/clusterhat/894a0d5f-6037-4531-bb40-44f65d3ca53d%40googlegroups.com.

[Michael von Blucher af Altona]

Thanks Peter

That worked.  My whole cluster is now on Docker version 19.03.1

I've obtained 2 more rpi 3b+  I'm going to build a stack, with the cluster controller as the head.

I've just got to figure out which raspbian image would be best for a starting point for the new pi's.

Already had a look at Hypriot OS. I found it had some annoying features.

Cheers

Michael B

To unsubscribe from this group and stop receiving emails from it, send an email to clust...@googlegroups.com.

[Peter J. Cross]

What features annoyed you?

Sent from my iPhone, please excuse any typos. 

On Oct 18, 2019, at 07:39, 'Michael von Blucher af Altona' via ClusterHAT <clust...@googlegroups.com> wrote:



To unsubscribe from this group and stop receiving emails from it, send an email to clusterhat+...@googlegroups.com.
To view this discussion on the web, visit https://groups.google.com/d/msgid/clusterhat/e74cac25-59da-49ce-a8d9-66ba901eddcf%40googlegroups.com.

[Michael von Blucher af Altona]

Hi Peter

Hypriot OS has the default login as "Pirate", which I didn't like, so I had to create a user "pi" and give all the same privilege as "Pirate".

This made the login consistent with the pi cluster.

It uses a specialized "cloud" configuration, which overrode every attempt I made to make my own networks settings. 

So persistence of changes becomes an issue.

I followed the instruction to allow persistence of changes, only to find that the instructions do not work.

No matter how I disabled the offending setting in the "cloud" config, my network settings disappeared on reboot.

I'll see if I can put in an override of my own tonight, to override the "cloud" configuration.

Failing that I might give Raspbian another try or maybe give Arrmbian a try.

Cheers

Michael B

To view this discussion on the web, visit https://groups.google.com/d/msgid/clusterhat/e74cac25-59da-49ce-a8d9-66ba901eddcf%40googlegroups.com.