Respect rate / data limit responses

yamen commented 3 years ago

Certain RPC endpoints will send 429 and 413 responses when rate or data limits have been hit. Currently serum-vial will retry in fixed 50ms backoff increments per subscribed market. This typically leads to hammering the server when it's already rate limiting.

Suggestions:

Increase and randomise reconnection time for 429 and 413 errors (eg between 250ms and 2500ms), and/or
Gate reconnection so only one market attempts to reconnect at a time

thaaddeus commented 3 years ago

Currently, I suggest to run own dedicated Solana RPC node, it's most stable this way, but I get your point regading public endpoints, pull request would be very welcomed.

thaaddeus commented 3 years ago

Hi, I've added exponential backoff strategy for RPC WS connections, hope it helps a bit.

yamen commented 3 years ago

Oh thanks a heap mate, that’s awesome. Sorry I did mean to pick this up but always so much harder to contribute to another project directly.

On Mon, 26 Jul 2021 at 8:48 pm, Thaddeus @.***> wrote:

Hi, I've added exponential backoff strategy for RPC WS connections, hope it helps a bit.

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/tardis-dev/serum-vial/issues/23#issuecomment-886591023, or unsubscribe https://github.com/notifications/unsubscribe-auth/AACVTZ7CXY7VZETJSHV7CILTZU4RPANCNFSM5AHZXXNQ .

-- Sent from Gmail Mobile

tardis-dev / serum-vial

Respect rate / data limit responses #23