State transition can take far too long if system clock is set to an earlier time

[jleveque]

If the system clock is set to an earlier time after a process is started but before it has entered the RUNNING state, supervisorctl start <process_name> can hang for a very long time waiting for the current system time to become greater than self.laststart + self.config.startsecs, depending on how far backward the system time has moved.

I have resolved the issue with the following modification to the transition() function in process.py:

def transition(self):
    now = time.time()
    state = self.state

    logger = self.config.options.logger

    # If system clock has moved backward, reset self.laststart to current system time
    if now < self.laststart:
        self.laststart = now;

   ...

This solution still ensures that the process has remained running for at least self.config.startsecs while only increasing the maximum wait time to 2 * self.config.startsecs if the system clock is detected to have moved backward.

As I'm not too familiar with this code, I'm not sure if this is the best solution, nor am I sure if any other functions may require a check of this nature as well, so I figured I would leave this suggestion here and let a Supervisor dev who is more familiar with the code base decide on the most appropriate solution.

Please let me know if I can be of further assistance.

[mnaberez]

What version of Supervisor are you using (supervisord --version)?

[jleveque]

Version 3.3.2

[jleveque]

To add more context, this is definitely a separate issue from #802. This is not related to the HTTP response issue that was apparently address this is a separate issue which can present similar symptoms.

[jleveque]

I pushed some changes to my PR (#1047) a few weeks ago in an attempt to address your comments. Please review the latest changes when you have a moment.