Enhancement: Adds Configurable First Retry Timeout for HTTP Calls #5050
Conversation
ananth7592
requested review from
kirankumarkolli,
FabianMeiswinkel,
sboshra and
Pilchie
as code owners
![github-actions[bot]](https://avatars.githubusercontent.com/in/15368?s=60&v=4){kind=small}
ananth7592
changed the title
Microsoft.Azure.Cosmos/src/HttpClient/HttpTimeoutPolicyControlPlaneRetriableHotPath.cs
Outdated
Show resolved
Hide resolved
Microsoft.Azure.Cosmos/src/HttpClient/HttpTimeoutPolicyControlPlaneRetriableHotPath.cs
Outdated
Show resolved
Hide resolved
ananth7592
force-pushed
the
ananth/add-timeout-retry
branch
2 times, most recently
from
522ecea to
c845e0e
Compare
Microsoft.Azure.Cosmos/src/HttpClient/HttpTimeoutPolicyControlPlaneRetriableHotPath.cs
Outdated
Show resolved
Hide resolved
| private HttpTimeoutPolicyControlPlaneRetriableHotPath(bool shouldThrow503OnTimeout) | ||
| { | ||
| this.shouldThrow503OnTimeout = shouldThrow503OnTimeout; | ||
| } | ||
|
|
||
| private readonly IReadOnlyList<(TimeSpan requestTimeout, TimeSpan delayForNextRequest)> TimeoutsAndDelays = new List<(TimeSpan requestTimeout, TimeSpan delayForNextRequest)>() | ||
| { | ||
| (TimeSpan.FromSeconds(.5), TimeSpan.Zero), | ||
| (TimeSpan.FromMilliseconds(Helpers.GetEnvironmentVariable( |
There was a problem hiding this comment.
If custom configuration it-self is >5seconds then is second retry with <== first worth?
There was a problem hiding this comment.
So if the customer chooses a value as high as 5 seconds, we wait for a timeout of 5 seconds the first time and 5 seconds the second time as well. That should be okay , right?
Also I am adding a minimum limit of 100ms to ensure the customer is not using a very low first timeout value.
There was a problem hiding this comment.
What's expected if they configure >5 seconds?
There was a problem hiding this comment.
Then , the first retry is attempted after 5 seconds (unless there is a cancellationToken with a lesser duration cancelling the task) .
Then if the HTTP call still fails, the second retry kicks in at the 10th (+5 seconds)
ananth7592
force-pushed
the
ananth/add-timeout-retry
branch
2 times, most recently
from
4887439 to
f4c738e
Compare
- Added support for configuring the first HTTP call timeout in the HTTP retry stack via an environment variable "AZURE_COSMOS_SDK_HTTP_FIRST_RETRY_TIMEOUT_VALUE_MS - Default first timeout remains 500ms, ensuring backward compatibility. - Provides customers with greater control over retry behavior to better suit their application needs.
ananth7592
force-pushed
the
ananth/add-timeout-retry
branch
from
f4c738e to
116b0c6
Compare
This update introduces the ability for customers to configure the first timeout value that the client waits before declaring an HTTP call as timed out. Once this timeout is reached, the exponential backoff retry strategy is triggered.
Changed Introduced
Example Customer use-cases
Increasing the first timeout (default: 500ms → custom value)
Useful when customers expect longer HTTP response times and want to allow the request more time to complete before triggering retries. This is particularly useful when the customer network topology is inherently causing their HTTP calls longer to materialize.
Reducing the first timeout (fail fast approach)
Helps customers fail fast by lowering the timeout below 500ms, ensuring that retries begin sooner and reducing overall request latency.
Pull Request Template
Description
Please include a summary of the change and which issue is fixed. Include samples if adding new API, and include relevant motivation and context. List any dependencies that are required for this change.
Type of change
Please delete options that are not relevant.
Closing issues
To automatically close an issue: closes #IssueNumber