Local ClusterODM vs Docker ClusterODM

I’m getting connection errors running local ClusterODM (launched w/ node index.js) versus the docker version. Runs fine with docker where I can statically scale nodes. Specifically - while uploading images with a single, local node it fails with:

Connection error: (‘Connection aborted.’, RemoteDisconnected(‘Remote end closed connection without response’))

I initially ran into this error while trying to use autoscaling so I tried it without the --asr parameter and I’m getting the same result. Maybe the non-docker version always tries autoscaling?

I’m running in AWS Linux 2, npm version 8.1, node version 16.13. I had difficulty getting the application to run until the node-libcurl problems documented in early posts were worked out.

I’ve grep’d through the libs directory of ClusterODM looking for this error and also run with inspector but have not been able to collect much debugging help or pinpoint where this error occurs.

Anyone else ever see this?

1 Like

with --debug on for WebODM, here’s the error/failure - ERROR FAILURE FOR Task [William Hilton Parkway - 8/12/2020] (e6d107b9-a2b4-4b65-b052-05e1c64b2d85): Connection error: (‘Connection aborted.’, RemoteDisconnected(‘Remote end closed connection without response’))
worker | [2021-11-22 15:54:06,964: ERROR/ForkPoolWorker-2] worker.tasks.process_task[bb261407-dc60-43d4-bddb-a25501658cb1]: FAILURE FOR Task [William Hilton Parkway - 8/12/2020] (e6d107b9-a2b4-4b65-b052-05e1c64b2d85): Connection error: (‘Connection aborted.’, RemoteDisconnected(‘Remote end closed connection without response’))

This occurs during uploading images to the processing node, gets roughly 7/8 done according to the progress bar, which takes forever, then this failure. Are there any ports this process is expecting to have available to ForkPoolWorker?

1 Like

Check your version of NodeJS (compare the version running in docker vs the version you’re running on your local machine). Time ago I’ve found that the very latest release of NodeJS breaks the current ClusterODM code (for reasons I have not yet understood).

1 Like

Correct, on other posts I confirmed that the node-libcurl dependency can’t run in the latest nodejs version (17.x) so I’ve dropped back to 16.x. This allowed me to successfully install and run the app. However, although ‘node index.js’ fires up fine and I can connect to it through WebODM as a processing node, it cannot seem to get past this connection error.

Not so with the docker version, which runs with 1 or many nodes attached to the cluster.

Ah ha, I see that container is running Node version 14.17.0. I’ll try reverting back to that.

1 Like

O.K. - so I reverted to node 14.17.0 and after reseting the node version with nvm and reinstalling I now have it running against the local ClusterODM just like the docker version!

Next stop - ASR!

2 Likes

Other docker neophytes like me might enjoy this reference - How To Get Information About Running Containers, Images In Docker? – POFTUT

2 Likes

Woo!

Celebration is in order!

2 Likes

This topic was automatically closed 30 days after the last reply. New replies are no longer allowed.