Skip to content

Check UDP errors after increasing IP ban duration from 2 minutes to 1 hour #28

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Closed
josecelano opened this issue Dec 19, 2024 · 2 comments

Comments

@josecelano
Copy link
Member

Relates to:

IP bans were increased from 120 seconds to 1 hour.

IPs are banned when they send more than 10 requests with the wrong connection ID.

You can see grafana dashboard before this change here.

image

The change was merged here at 2024-12-19 12:26 UTC.

We were banning 100 req/sec on average. I will run the tracker with the new configuration for hours to compare it with the previous data.

@josecelano
Copy link
Member Author

After 12 hours of running the 1-hour ban:

image

There are fewer UDP errors as I expected. However, I don't know why the connection requests graph seems to follow error graphs. Why are connections decreasing when we ban IPs because they send wrong connection IDs? That means clients make a connection request but do not send the right one in the following announce/scrape requests. But error since to be equally distributed between:

  • cookie value is expired
  • cookie value is from future

If the problem is that clients are reusing the same connection ID for too long, we will only have cookie value is expired errors.

Maybe after running this experiment #29 we can know more.

I have also analysed the behaviour of one banned client. That client was making the connect request but then it was always using the same connection ID:

ConnectionId(I64(-2325979768337461225))

But I don't remember if there was an announcement request returning that ID. I don't think so or it was made before the date range I got from logs.

@josecelano
Copy link
Member Author

The good thing is apparently the number of announce requests handled on average has increased a little bit. Maybe because we are not wasting resources sending the error response to "bad" clients.

cc @da2ce7

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

1 participant