Add status codes configuration #96
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
schneems
added a commit
to schneems/platform-api
that referenced
this pull request
Nov 22, 2019
By default when you get a rate-limit event while using the platform-api the method call will fail. Since there's no feedback built into the client, another call might be attempted (for example if the API is being hit via a background worker). This is bad for Heroku because it means we now are seeing a stream of requests that will never complete (because of the rate limit) and it is bad for the end user because they have a flurry of errors that are unexpected and unhandled. This PR builds on top of the project https://github.com/schneems/rate-limit-gcra-client-demo to automatically find a value that the client can sleep for so that it can maximize throughput while still minimizing the number of requests that are rate limited. At a high level when it sees a 429 response from the server, it will sleep and then retry the response. If it gets another 429 then the sleep amount will be multiplicatively increased until it can make a successful request. As the client is able to make successful requests the amount of sleep time is reduced by a subtractive amount based off of the current number of requests allowed (as reported by the server), the amount of time since the last rate limit event, and it's current value. This logic somewhat mirrors TCP "slow start" behavior (though in reverse). In simulations over time we end up seeing from 2-3% of requests rate limited.  > Graph from README of the simulation readme This PR has been built on the work of other changes added to heroics: - interagent/heroics#95 - interagent/heroics#96 ## Discussion In addition to the implementation, the one last unknown is what the default logging behavior should be. While rate-throttling by default provides a good experience, we need to provide feedback to the user letting them know that it is happening. This PR was paired with @lolaodelola
schneems
pushed a commit
to schneems/platform-api
that referenced
this pull request
Nov 22, 2019
By default when you get a rate-limit event while using the platform-api the method call will fail. Since there's no feedback built into the client, another call might be attempted (for example if the API is being hit via a background worker). This is bad for Heroku because it means we now are seeing a stream of requests that will never complete (because of the rate limit) and it is bad for the end user because they have a flurry of errors that are unexpected and unhandled. This PR builds on top of the project https://github.com/schneems/rate-limit-gcra-client-demo to automatically find a value that the client can sleep for so that it can maximize throughput while still minimizing the number of requests that are rate limited. At a high level when it sees a 429 response from the server, it will sleep and then retry the response. If it gets another 429 then the sleep amount will be multiplicatively increased until it can make a successful request. As the client is able to make successful requests the amount of sleep time is reduced by a subtractive amount based off of the current number of requests allowed (as reported by the server), the amount of time since the last rate limit event, and it's current value. This logic somewhat mirrors TCP "slow start" behavior (though in reverse). In simulations over time we end up seeing from 2-3% of requests rate limited.  > Graph from README of the simulation readme This PR has been built on the work of other changes added to heroics: - interagent/heroics#95 - interagent/heroics#96 ## Discussion In addition to the implementation, the one last unknown is what the default logging behavior should be. While rate-throttling by default provides a good experience, we need to provide feedback to the user letting them know that it is happening. This PR was paired with @lolaodelola
schneems
pushed a commit
to heroku/platform-api
that referenced
this pull request
Apr 20, 2020
By default when you get a rate-limit event while using the platform-api the method call will fail. Since there's no feedback built into the client, another call might be attempted (for example if the API is being hit via a background worker). This is bad for Heroku because it means we now are seeing a stream of requests that will never complete (because of the rate limit) and it is bad for the end user because they have a flurry of errors that are unexpected and unhandled. This PR builds on top of the project https://github.com/schneems/rate-limit-gcra-client-demo to automatically find a value that the client can sleep for so that it can maximize throughput while still minimizing the number of requests that are rate limited. At a high level when it sees a 429 response from the server, it will sleep and then retry the response. If it gets another 429 then the sleep amount will be multiplicatively increased until it can make a successful request. As the client is able to make successful requests the amount of sleep time is reduced by a subtractive amount based off of the current number of requests allowed (as reported by the server), the amount of time since the last rate limit event, and it's current value. This logic somewhat mirrors TCP "slow start" behavior (though in reverse). In simulations over time we end up seeing from 2-3% of requests rate limited.  > Graph from README of the simulation readme This PR has been built on the work of other changes added to heroics: - interagent/heroics#95 - interagent/heroics#96 ## Discussion In addition to the implementation, the one last unknown is what the default logging behavior should be. While rate-throttling by default provides a good experience, we need to provide feedback to the user letting them know that it is happening. This PR was paired with @lolaodelola
schneems
pushed a commit
to heroku/platform-api
that referenced
this pull request
May 4, 2020
By default when you get a rate-limit event while using the platform-api the method call will fail. Since there's no feedback built into the client, another call might be attempted (for example if the API is being hit via a background worker). This is bad for Heroku because it means we now are seeing a stream of requests that will never complete (because of the rate limit) and it is bad for the end user because they have a flurry of errors that are unexpected and unhandled. This PR builds on top of the project https://github.com/schneems/rate-limit-gcra-client-demo to automatically find a value that the client can sleep for so that it can maximize throughput while still minimizing the number of requests that are rate limited. At a high level when it sees a 429 response from the server, it will sleep and then retry the response. If it gets another 429 then the sleep amount will be multiplicatively increased until it can make a successful request. As the client is able to make successful requests the amount of sleep time is reduced by a subtractive amount based off of the current number of requests allowed (as reported by the server), the amount of time since the last rate limit event, and it's current value. This logic somewhat mirrors TCP "slow start" behavior (though in reverse). In simulations over time we end up seeing from 2-3% of requests rate limited.  > Graph from README of the simulation readme This PR has been built on the work of other changes added to heroics: - interagent/heroics#95 - interagent/heroics#96 ## Discussion In addition to the implementation, the one last unknown is what the default logging behavior should be. While rate-throttling by default provides a good experience, we need to provide feedback to the user letting them know that it is happening. This PR was paired with @lolaodelola
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
When we're expecting the server to rate limit our client, then a different status code may be acceptable such as a 429. This PR introduces the ability to add additional acceptable status codes.