Add exponential retry logic if Github API returns 500

internal/github/github.go

@@ -2,6 +2,9 @@ package github
 
 import (
 	"context"
+	"errors"
+	"fmt"
+	"time"
 
 	"github.com/google/go-github/v38/github"
 	"golang.org/x/oauth2"
@@ -27,6 +30,8 @@ type Client interface {
 
 type client struct {
 	ghc *github.Client
+	maxRetries int
+	retryDelay time.Duration
 }
 
 func NewClient(ctx context.Context, token string) Client {
@@ -36,13 +41,89 @@ func NewClient(ctx context.Context, token string) Client {
 				AccessToken: token,
 			},
 		))),
+		maxRetries: 5,
+		retryDelay: 1 * time.Second,
 	}
 }
 
 func (c *client) GetCombinedStatus(ctx context.Context, owner, repo, ref string, opts *ListOptions) (*CombinedStatus, *Response, error) {
-	return c.ghc.Repositories.GetCombinedStatus(ctx, owner, repo, ref, opts)
+	var statusResp *CombinedStatus
+	var resp *Response
+	var err error
+
+	for attempt := 0; attempt < c.maxRetries; attempt++ {
+		statusResp, resp, err = c.ghc.Repositories.GetCombinedStatus(ctx, owner, repo, ref, opts)
+		if err == nil {
+			return statusResp, resp, nil
+		}
+
+		// Check if context is canceled or deadline exceeded before retrying
+		if ctx.Err() != nil {
+			return nil, resp, fmt.Errorf("context error while getting combined status: %w", ctx.Err())
+		}
+
+		// Don't retry on deadline exceeded errors
+		if errors.Is(err, context.DeadlineExceeded) {
+			return nil, resp, err
+		}
+
+		// Only retry on 5xx server errors
+		if resp != nil && (resp.StatusCode < 500 || resp.StatusCode > 599) {
+			return statusResp, resp, err
+		}
+
+		// Wait with exponential backoff before retrying
+		if attempt < c.maxRetries-1 {
+			backoffDuration := c.retryDelay * time.Duration(1<<attempt)
+			select {
+			case <-ctx.Done():
+				return nil, resp, ctx.Err()
+			case <-time.After(backoffDuration):
+				// Continue with retry
+			}
+		}
+	}
+
+	return nil, resp, fmt.Errorf("failed to get combined status after %d retries: %w", c.maxRetries, err)
 }
 
 func (c *client) ListCheckRunsForRef(ctx context.Context, owner, repo, ref string, opts *ListCheckRunsOptions) (*ListCheckRunsResults, *Response, error) {
-	return c.ghc.Checks.ListCheckRunsForRef(ctx, owner, repo, ref, opts)
+	var checksResp *ListCheckRunsResults
+	var resp *Response
+	var err error
+
+	for attempt := 0; attempt < c.maxRetries; attempt++ {
+		checksResp, resp, err = c.ghc.Checks.ListCheckRunsForRef(ctx, owner, repo, ref, opts)
+		if err == nil {
+			return checksResp, resp, nil
+		}
+
+		// Check if context is canceled or deadline exceeded before retrying
+		if ctx.Err() != nil {
+			return nil, resp, fmt.Errorf("context error while listing check runs: %w", ctx.Err())
+		}
+
+		// Don't retry on deadline exceeded errors
+		if errors.Is(err, context.DeadlineExceeded) {
+			return nil, resp, err
+		}
+
+		// Only retry on 5xx server errors
+		if resp != nil && (resp.StatusCode < 500 || resp.StatusCode > 599) {
+			return checksResp, resp, err
+		}
+
+		// Wait with exponential backoff before retrying
+		if attempt < c.maxRetries-1 {
+			backoffDuration := c.retryDelay * time.Duration(1<<attempt)
+			select {
+			case <-ctx.Done():
+				return nil, resp, ctx.Err()
+			case <-time.After(backoffDuration):
+				// Continue with retry
+			}
+		}
+	}
+
+	return nil, resp, fmt.Errorf("failed to list check runs after %d retries: %w", c.maxRetries, err)
 }