adrtsc
commented
1 year ago I found in the end that it was not an issue with the meta file put with the args_cellpose_segmentation.json. The mistake that triggered the json decoder error was the following:
fr{
"level": 2,
"wavelength_id": "A01_C01",
"output_label_name": "nuclei",
"input_ROI_table": "FOV_ROI_table",
"model_type": "nuclei",
"diameter_level0": 120
}I tried resubmitting the workflow with the erroneous args_cellpose_segmentation.json and I only got the json decoder error this time and the database was not locked. So it might have been a weird coincidence that it happened the first time.
jluethi
tcompa
The database on the UZH deployment using server 1.2.2 became locked today. After Adrian ran into an error when providing a wrong json file during adding a task to a workflow (likely an error in the meta json file), the database became locked and any user using fractal client would see this error:
The reported original error in the workflow was:
Though I'm a bit surprised with that. I've seen this error multiple times when adding wrong task argument jsons and it never froze the server.
Are server logs written to disk anywhere? The scroll-back in the screen only revealed this type of error over & over for different requests by users to do any operation (even just
fractal user whoami), but not the original error.This is the error that gets triggered every time a user tries to interact with the server via the client:
Restarting the server solved the issue, so it seems some process was not shut down correctly after an exception.
tl;dr: Let's keep an eye on this. Not sure how we can reproduce the error and whether it's another case of something that would get better with postgres. We should try adding a task to a workflow with a corrupt meta.json file (cc @adrtsc do you still have the one that caused the error for you?)