I already has one change which silences a good deal of floods:
--- a/src/utils/net/ip.ml
+++ b/src/utils/net/ip.ml
@@ -360,7 +360,7 @@ let _ =
(* lprintf "Ip found for %s: %s\n" job.name (to_string ip); *)
job.handler ip
end else begin
- lprintf_nl (_b "[DNS] could not resolve %s, check URL") job.name;
+ (* lprintf_nl (_b "[DNS] could not resolve %s, check URL") job.name; *)
job.err_handler ();
raise Not_found
end
(I am sure that's wrong - the correct solution is to not look up some hosts repeatedly, rather than simply silencing failures)
After about 90 minutes, it is collecting a few:
$ cut -f 3-6 -d ' ' mlnet.log | sort | uniq -d -c
...
475 [BT] udpt interact with
...
294 [HTTPcl] 404: Not found
98 [HTTPcl] 409: bad reply
I see most of them are probably outdated/invalid trackers... nonetheless, it might be a good idea to back off on unresponsive trackers, rather than repeatedly looking them up?
I already has one change which silences a good deal of floods:
(I am sure that's wrong - the correct solution is to not look up some hosts repeatedly, rather than simply silencing failures)
After about 90 minutes, it is collecting a few:
I see most of them are probably outdated/invalid trackers... nonetheless, it might be a good idea to back off on unresponsive trackers, rather than repeatedly looking them up?