With requests and urllib3 under the hood, the Cortex Data Lake Python SDK offers access to a variety of powerful features, adapters and design patterns, that developers can use to extend the base functionality. The following tutorial presents a recipe for overridding the default
Retry strategy in order to handle situations where API rate limits are exceeded.
The Query Service API currently implements a sliding window rate limit of 5,000 queries/hour.
Before we begin, it's important to note that the base CDL Python SDK
HTTPClient class implements a general case transport adapter, configured with the following settings:
Although good enough for the majority of cases, there are situations where you could implement a retry strategy to handle specific error conditions. One such case is when the Query Service API returns an HTTP status 429 code, also known as a "Too Many Requests" error. This can occur when your client exceeds the Query Service API rate limit of 5,000 queries/hour.
With the default
HTTPClient, you'd be forced to code your own retry strategy, complete with a retry interval and backoff algorithm.
As you might have guessed by now, there's a way to configure the CDL Python SDK to automatically retry after receiving an HTTP
429 response code ("Too Many Requests" error). Moreover, we can define a retry strategy that also implements exponential backoff, all with a few lines of code!
So let's break down what's happening here:
- Import the
- Define a retry strategy to handle HTTP
- Pass the
QueryService()class constructor using the
Now, if we exceed the Query Service API rate limit, the CDL Python SDK will automatically implement the retry strategy with a retry interval of
0s, 14s, 28s, 56s, 112s, 224s, 448s, 896s, 1792s, 3584s, more than enough time for the rate limiting window to reset.
Retry can work in conjunction with
timeout, connection/read timeouts can only be applied on a per-request basis.
For more information see the official reference documentation for the