[18.0] Job batches stuck in In Progress

[stamtos]

Module

queue_job_batch

Describe the bug

Job batches stuck in In Progress even when all are finished.

To Reproduce

Odoo 18.0 Enterprise

Steps to reproduce the behavior:

1. Create a few Batch Jobs with more than 1 jobs.
2. Wait a few hours. and randomly some batches will stuck in in progress

Expected behavior The batch should marked as finished when all jobs are finished

Additional context

[amh-mw]

Can you provide additional details, such as logs? I have encountered something like this in the past when my jobs took longer than --limit-time-* ^1 settings allowed.

[stamtos]

Hi @amh-mw, thank you for your response. I will increase the limit-time-* values and see if this helps.

I looked at the jobs in another batch, and all of them finished in the "done" state while taking less than the default limit-time-cpu value of 60 seconds. Additionally, the total time for all jobs was less than 60 seconds. So I do not think it is because of a timeout.

I do not see any warning or error logs.

here is my job pseudocode:

def process(self, param):
    _logger.info(f"Processing {len(self)} items with parameter {param}")
    url = self.env['ir.config_parameter'].sudo().get_param('service.url')  
    api_key = self.env['ir.config_parameter'].sudo().get_param('service.key')
    for rec in self:
        headers = {
            "Content-Type": "application/json",
            "Authorization": f"Bearer {api_key}"
        }
        input_text = rec.name  
        data = {"input": input_text}  
        
        try:
            response = requests.post(url, headers=headers, json=data)
            response.raise_for_status()
            _logger.info(f"Response: {response.json()}")
            output = response.json().get("result", {})  
            
            value1 = output.get("key1", rec.value1)
            value2 = output.get("key2", rec.value2)
            
            data_dict = {
                'key1': value1,
                'key2': value2
            }
            
            rec.update_values('field', data_dict)
            
        except BaseException as e:
            _logger.error(f"Failed for input '{input_text}': {e}")
            raise RetryableJobError(f"Failed for input '{input_text}': {e}")

And here is the batch method:

def start(self):
    batch = self.env['queue.job.batch'].sudo().get_new_batch('Batch Process')  
    batch_size = 1  # Adjust as needed
    for i in range(0, len(self), batch_size):
        subset = self[i:i + batch_size]
        subset.sudo().with_context(job_batch=batch).with_delay(
            channel='root.process',
            description='Process subset',  
        ).process('param_value') 
    return True

PS: also tried adjusting the batch_size from 10 to 1

[stamtos]

found the root cause of this issue. a race condition in the batch state recompute when multiple jobs finish at the same time (and using at least 2 channels). This leaves some batches stuck in “In Progress” even though all jobs are “Done.”

To validate my hypothesis, I will implement some fixes in my codebase. After testing and validating them in my deployment environment, if this really fixes the problem, I will open a pull request with the changes.

https://github.com/OCA/queue/blob/2e0ed2cc9022eecc517315308ce25c87e0c214ec/queue_job_batch/models/queue_job.py#L22-L32

another example:

[remi-filament]

Hi @stamtos any news on that one ? We face the same problem here, if your code solves the issue, would be interested to know / get a PR for that !

Thanks !

[stamtos]

@remi-filament, due to other work priorities, I haven't tested anything yet. As a workaround, we have a cron running check_state() periodically:

env['queue.job.batch'].search([('state', '=', 'progress')]).check_state()
env.cr.commit()

Hope it helps in the meantime.