shared client: Fix global timeout not being respected

This fixes a bug where the shared client did not respect `c.Client.Timeout`, resulting in DNS requests timing out too soon. A per Go docs, `Client.Timeout` is supposed to also apply to the write timeout. This is implemented in the regular client by using `Client.getTimeoutForRequest` to determine the actual timeout. This commit adapts the shared client code to do the same. Signed-off-by: Sebastian Wicki <[email protected]>
cilium · Apr 16, 2024 · d47d0dd · d47d0dd
1 parent 4e6b438
commit d47d0dd
Showing 1 changed file with 1 addition and 1 deletion.
diff --git a/shared_client.go b/shared_client.go
@@ -291,7 +291,7 @@ func (c *SharedClient) ExchangeSharedContext(ctx context.Context, m *Msg) (r *Ms
 
 	// This request keeps 'c.requests' open; sending a request may hang indefinitely if
 	// the handler happens to quit at the same time. Use ctx.Done to avoid this.
-	timeout := c.Client.writeTimeout()
+	timeout := c.getTimeoutForRequest(c.Client.writeTimeout())
 	ctx, cancel := context.WithTimeout(ctx, timeout)
 	defer cancel()
 	respCh := make(chan sharedClientResponse)