Problems with collecting coverage data for the uwsgi framework

[WeiMengXS]

Error:

2025-08-20 16:41:58.176397 : exception Couldn't use data file '/usr/local/services/xxxxxx/.coverage.cvm-api-0.112.XIvCOBix': no such table: context , Traceback (most recent call last)

I'm using uwsgi to start my service, and a thread is created to collect and report coverage data:

def _cover_report(_cover):
    reporter = Reporter(_cover, "{token}", "{server}", "{project_name}", "{raw_name}", "{commit}", trim_path,"{user_agent}",{flags})
    while True:
            try:
                wlog('whiletrue  %s , %s ' % (os.getpid(), {interval}))
                time.sleep({interval})
                result = reporter.run()
                _cover.erase()
                wlog(result)
            except Exception as e:
                msg = traceback.format_exc()
                wlog('exception %s , %s' % (e, msg))
                time.sleep({interval})


qcov_origin_application = {entrance_func}
def {entrance_func}(environ=None, start_response=None):
    cover.start()
    res =  qcov_origin_application(environ, start_response)
    global report_thread_started
    if not report_thread_started:
        thread = threading.Thread(target=_cover_report, args=(cover,))
        thread.setDaemon(True)
        thread.start()
        report_thread_started = True
    return res

It seems that the erase() step causes the database file of other processes to be deleted.

My expectation is that each process collects its own data and then reports them to my server independently.

Please give me some guidance！

[nedbat]

I don't understand everything you are doing here, but if you want separate processes to all collect data, you should use parallel=True and you don't need erase() at all. If you are still having trouble, provide a complete runnable example so we can reproduce the problem.

[WeiMengXS]

Hello,

My current scenario is that I instrument other services and periodically collect coverage data from coverage, then report it to provide e2e coverage monitoring.

When parallel=True, does a "single process" have the collected data from all processes, or is the data collected separately in each process?

Actually, I have enabled parallel=True, but merging(combine()) the coverage files can be time-consuming, so I perform the merge operation on the server side. However, the memory usage of each process is still quite high, and _analyze is also slow. So I would like to know: after uploading the data, is it possible to clean up the memory or clear the current process's SQLite (coverage data)?

When I try to clean up (for example, by calling erase()), I encounter the following error when trying to call cover.get_data() again:

2025-08-20 16:41:58.176397 : exception Couldn't use data file '/usr/local/services/xxxxxx/.coverage.cvm-api-0.112.XIvCOBix': no such table: context , Traceback (most recent call last)

How can I safely clean up or reset the coverage data for the current process after uploading, without causing this error?

[nedbat]

If you want to continue to collect data you could create a new Coverage object.

[WeiMengXS]

Thank you very much for your reply.

Are you saying to call start() after erase()?

But this approach seems problematic for managing the timers in other processes.

With coverage, is there a way to handle this issue?

Because between erase() and start(), other processes might be calling get_data().

Are there any similar best practices for my scenario? Or how to use it would be more reasonable?

[nedbat]

Don't use erase() at all.