Intel NUC 2.26+rev1 not downloading production application

I am trying to install my application on our production devices (using production image).
Everytime it falls back to 0% after reaching somewhere near 15%. I had this problem before and it was fixed using a newer supervisor version by the support, but it was a HTTP 500 code. This time I am not getting an error, it just tries to re-download. Surprisingly the development image downloads. What might be the problem? I also made a test application using the production version on the same device. It was smaller of size, failed sometimes but managed to get to 100%. Thanks

07.12.18 15:49:34 (+0100) Creating network 'default'
07.12.18 15:49:34 (+0100) Creating volume 'resin-data'
07.12.18 15:49:34 (+0100) Downloading image ''
07.12.18 16:07:50 (+0100) Downloading image ''
07.12.18 16:25:34 (+0100) Downloading image ''
07.12.18 16:40:36 (+0100) Downloading image ''

Hey @rquant
Those logs would normally point to a poor network connection but I’d be happy to look at the device for you.

Can you provide me with the dashboard link and enable support access please?

Hello, support access granted.
I now switched to the office internet 150 Mbit/s and now get the same with HTTP 500, which were caused by the Supervisor Version. Update to 8.3.X (which is not released yet) helped resolve the error. I have to setup 20 production devices, but am unable to download the application. What do you recommend me to do?

Hey @rquant so I watched the device that you linked. It’s hitting a bug related to OOM killings of the supervisor. I would recommend you use the latest version of balenaOS, which does not suffer from this issue, especially for your productions device that are yet to be setup.

I am using the latest balenaOS. 2.26+rev1 production and 8.0.0 supervisor. Would it help to increase the RAM size?

Apologies, you’re correct, the latest version in staging has this, but it hasn’t quite finished internal testing, ready to reach production. This should be completed within the week, if not slightly longer. Does this work for you?

That would break our deadlines with our customer. Latest would be next week monday for final testing. We now build the whole application assuming it would work with the production image. I remember during development it downloaded, maybe something changed with the versions.

Would it be possible to update 5 devices manually to the newest supervisor?

Hi @rquant, could you share the device uuids via private message? Cheers!

@CameronDiver We are having the same issue I believe. In your message I read that this should have been fixed in 2.27 and that should be released by now. I don’t think it is… is it? Do you have a new timeframe?

1 Like

I think, he is responsible for updates. We are also waiting. Using 2.13, which is more stable than 2.26

Thanks @rquant for the reply. We’ll have to try 2.13 then, because using the recommended 2.26 basically makes the platform useless because you can’t deploy…

Hi @Guuz and @rquant we have released OS 2.29 on the intel nuc now and this has the fix for supervisor where it would be starved from memory.


hi @shaunmulligan,

Hi @rquant, it seems there is a problem with OS downloads on . Are you using the unmanaged OS ? On balenaCloud the image download is working. We are looking into it currently.

On the question of how the release happens after merge. Once it gets merged on meta-balena, it then needs to get PR’d to each device specific repo, in the intel-nuc this is the repo: . Once that PR is merged, then a deploy is kicked off by our devices team, which builds and puts the image on our staging environment. From there we run through a long set of tests to validate the release. Once our QA process passes, then we deploy to production .

@rquant the 2.29 OS version download should be working correctly now.