Anyway to set sleep timer for --low-level-retries ?

To my knowledge, no as the sleep backoff uses an exponential algorithm as it increases based on the number of failures in a row.

@ncw would be able to shed some light if something changed recently to alter that statement.

1 Like

The problem that I have having is the algorithm is not doing a good job in my case. It takes 5-6 retries, so if I make it wait 5s or 10s that may solve problem.

Using --tpslimit=1 does not make any difference.

What backend are you using?

Some of the backends have an exponential retry schedule you must follow and some don't.

Do you want it to sleep longer or shorter?

Sleep longer. Using OneDrive.

This is the config from the onedrive backend

        minSleep                    = 10 * time.Millisecond
        maxSleep                    = 2 * time.Second
        decayConstant               = 2 // bigger for slower decay, exponential

So the max sleep between retries is 2 seconds at the moment

You could try increasing --low-level-retries?

Yeah the low-level-retries starts with 10ms then increment with 2x just like you coded. But the problem is it's causing too much HTTP calls which I am trying to avoid, Microsoft made this very clear on their docs.

Yeah, using 30 but that does not stop the HTTP calls.

I think rclone does all those things

  • it implements (truncated) exponential backoff
  • it obeys Retry-After headers if sent

It might be that it isn't tuned correctly - the max sleep maybe needs to be higher than 2 seconds?

Try modding the code as I pointed out in the issue and see if you can get it to behave how you want it.

This topic was automatically closed 60 days after the last reply. New replies are no longer allowed.