ServiceListenTimeout

[scottschmidl]

Describe the problem

I am using fiftyone to download open-images-v7 to also fine tune a YOLO model. Fiftyone installs/import fine, but when I download and load the images I'm getting an error message fiftyone.core.service.ServiceListenTimeout: fiftyone.core.service.DatabaseService failed to bind to port. NOTE: I'm using the package manager UV: https://astral.sh/blog/uv. I was using conda (and if i recall it was working fine with conda) but my company has banned it.

powershell -ExecutionPolicy ByPass -c "irm https://astral.sh/uv/install.ps1 | iex"
uv self update
uv venv --python 3.11
.venv/Scripts/activate
uv pip install -r requirements.txt: requirements.txt
uv pip list: uvpiplist.txt

NOTE: this error is happening after i run the initial download which also fails after some time, but downloads some images.

Code to reproduce issue

download.txt

System information

• OS Platform and Distribution (e.g., Linux Ubuntu 22.04): Microsoft Windows 11 Enterprise Version 10.0.22631 Build 22631
• Python version (python --version): 3.11.11
• FiftyOne version (fiftyone --version): FiftyOne v1.3.0, Voxel51, Inc.
• FiftyOne installed from (pip or source): pip, but i used UV.

Other info/logs

LOGS.txt

mongo.log

I am opening up this issues because I have spent hours googling and using GenAI to help troubleshoot, but I can't find a resolution.

Willingness to contribute

The FiftyOne Community encourages bug fix contributions. Would you or another
member of your organization be willing to contribute a fix for this bug to the
FiftyOne codebase?

• Yes. I can contribute a fix for this bug independently
• [] Yes. I would be willing to contribute a fix for this bug with guidance
  from the FiftyOne community
• [] No. I cannot contribute a bug fix at this time

[benjaminpkane]

Hi @scottschmidl. Configuring your own MongdoDB connection may fix your issue. I will have to double check the current state of builtin DB support on Windows this week

[benjaminpkane]

Ah, I see you have provided the mongo.log. This is feature compatibility version issue. This link will help fix the issue.

Was your database FCV set to 6.0 at some point? Maybe FIFTYONE_DATABASE_URI was set and then unset? This` is not something FiftyOne does currently, although it will be handled automatically in soon #5441

[scottschmidl]

I'm not sure about FCV. I didn't make any change until after this issue started happening. I don't understand, I wasn't having this issue with conda which I was previously using on the same windows computer.

Just to be clear, you're saying that I need to configure MongoDB now? Just FYI, I am just trying to download these images and then use YOLO to fine-tune my model, but this is for work. I'm a MLE and this is a huge inconvenience. I really liked the fiftyone interface and functionality, but I might have to just remove it and change course because I have deadlines.

[benjaminpkane]

If you don't care about persistent datasets you can delete the existing database. Feel free to do what you will

rm -rf ~/.fiftyone/var/lib/mongo

[scottschmidl]

Just curious will that fix my issue?

Since I'm using a YOLO model, I saw that I can download the images directly through Ultralytics, except that they use fiftyone to download the images before training the model. Ultralytics is now giving me this boto error from fiftyone: botocore.httpchecksum - INFO - Skipping checksum validation. Response did not contain one of the following algorithms: ['crc32', 'sha1', 'sha256'].

I tried pip uninstall > pip install but it didn't fix the issue

[benjaminpkane]

Just curious will that fix my issue?

Yes, I believe it will. It will clear the database so the next startup can initialize the FCV correctly.

Since I'm using a YOLO model, I saw that I can download the images directly through Ultralytics, except that they use fiftyone to download the images before training the model. Ultralytics is now giving me this boto error from fiftyone: botocore.httpchecksum - INFO - Skipping checksum validation. Response did not contain one of the following algorithms: ['crc32', 'sha1', 'sha256'].

I tried pip uninstall > pip install but it didn't fix the issue

That is an INFO log, so I don't believe it should cause any errors. @ehofesmann have you seen this before?

[ehofesmann]

I have not seen that checksum log before, but I agree that given it's only an INFO log, I would assume it's not actually causing errors. If there are any downstream issues you're seeing from Ultralytics, please share any additional logs or error tracebacks you have from that.

[afoley587]

For your MongoDB issue, did you recently do an upgrade of fiftyone? If so, can you let us know from which version to which version?

Looking at your mongoDB log, it looks like you pip install fiftyone with an older version of fiftyone. The older version was packaged with mongoDB 4.4. The conda cache probably never upgraded it because the version isn't pinned in your requirements.txt.

Then, when uv was used instead of conda, a newer version of fiftyone was used. This bumped the mongoDB version to 6.0+. Upgrades of more than one major version isn't supported by mongo, and I believe that's what the error is.

But, if you know which version you upgraded from, we might be able to provide you an upgrade path to address this.