okx rate-limit incorrect

• OS: linux
• Programming Language version: python
• CCXT version: 1.87.54

When using asyncio to retrieve a lot of leverage tiers - it becomes apparent that the rate-limit for that endpoint is incorrect. In my understanding, ccxt’s throttler should kick in based on configured rate-limits - and not make the requests to the exchange too often.

based on the below - i’d expect no output other than the “128” for the length of pairs.

import ccxt.async_support as ccxt_async
exchange = ccxt_async.okx({'options': {'defaultType': 'swap'}, 'enableRateLimit': True})

import asyncio
async def fetch_levt(p):
    try:
        return await exchange.fetch_market_leverage_tiers(p)
    except Exception as e:
        print(p)
        print(e)
        return None

async def test():
    lt = []
    m = await exchange.load_markets()
    for p in m.keys():
        if p.endswith(':USDT'):
            lt.append(fetch_levt(p))
    print(len(lt))
    await asyncio.gather(*lt, return_exceptions=True)

asyncio.run(test())

128
1INCH/USDT:USDT
okx {"msg":"Requests too frequent.","code":"50011"}
BSV/USDT:USDT
okx {"msg":"Requests too frequent.","code":"50011"}
ENJ/USDT:USDT
okx {"msg":"Requests too frequent.","code":"50011"}
ICP/USDT:USDT
okx {"msg":"Requests too frequent.","code":"50011"}
NFT/USDT:USDT
okx {"msg":"Requests too frequent.","code":"50011"}
SAND/USDT:USDT
okx {"msg":"Requests too frequent.","code":"50011"}
THETA/USDT:USDT
okx {"msg":"Requests too frequent.","code":"50011"}
XLM/USDT:USDT
okx {"msg":"Requests too frequent.","code":"50011"}

there are however quite some okx {"msg":"Requests too frequent.","code":"50011"} messages mixed in - causing a delay in operations.

Now obviously we can handle this and retry (freqtrade does so) - but it’s still slowing us down more than it should. It doesn’t seem “too far” off - but just slightly - enough to cause an exception on 8 out of 128 requests - so roughly one every 16 requests.

Based on the docs - the rate-limit for this endpoint should be Rate Limit: 10 requests per 2 seconds … it’s currently set to “2” - but it’s unclear to me what unit “2” is in. (https://github.com/ccxt/ccxt/blob/bbda875a9de4209535afe6987a14534204dcae42/js/okx.js#L170)

[xmatthias]

@frosty00 thanks, you’re right, i can no longer reproduce this - like - not at all from different locations. not with the version i highlighted the issue in the first place - so it’s not a ccxt change that caused this.

I suspect something on okx’s side changed (maybe they’re now less strict - or they fixed a bug on their side) - either way, i’ve now tried for 2 days from different locations / systems without ever running into this problem again, so i think this can be closed - as there’s nothing actionable.

[frosty00]

What your experiencing is likely a known issue with ccxt asyncio in python, and it is being worked on to be fixed asap, due to the leaky bucket algorithm, which is implemented in the throttler, sometimes it sends an extra request or two, and some exchanges are extremely strict of this, and others are not.

Actually there is no issue here, it’s working as intended. It has the capacity for an extra request which allows markets to load faster