The main takeaways are to do exponential backoff and add some random jitter. Do that and the recovery will go more quickly. O.k. so what am I tweeting about?
-
-
If you provide an API, provide a great client that implements good client behavior. Most bad API consumers are busy, not malicious. Make doing the right thing the easiest path.
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
this is sort of the point of envoy et Al, no? rather than distributing retry implementation across a thousand snowflake clients you source it to a transparent proxy. teams still have to set this effectively but at least networking admins can audit and complain
-
this can be extended to be a policy based networking solution eg istio which for many teams solves this problem mostly
End of conversation
New conversation -
-
-
Truly interesting! But if this is everyone's issue, shouldn't it be on the protocol instead of clients? I remember FT CORBA, is it time for FT HTTPS?
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
@threadreaderapp Kindly unroll -
Hello please find the unroll here: Thread by
@colmmacc: "Rare evening technical thread: about distributed systems, exponential backoff, and the tragedy of the commons problem. L […]" https://threadreaderapp.com/thread/1090453892687523847.html … Share this if you think it's interesting.
End of conversation
New conversation -
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.