What causes "body close failed at %s" in Netflix/chaosmonkey?

Read the full body (ioutil.ReadAll) before the deferred close runs.

body close failed at %s

Error message

body close failed at %s

What it means

This error is produced by the deferred cleanup in Spinnaker.GetInstanceIDs when resp.Body.Close() returns an error after a successful HTTP GET. The library wraps it with the URL so you know which response body could not be closed. Note a quirk: errors.Wrapf(err, ...) wraps the outer err (nil at that point conceptually) rather than cerr, so the close error itself can be lost — but the message still flags that closing failed.

Source

Thrown at spinnaker/spinnaker.go:264

			continue
		}

		c <- app
	}
}

// GetInstanceIDs gets the instance ids for a cluster
func (s Spinnaker) GetInstanceIDs(app string, account D.AccountName, cloudProvider string, region D.RegionName, cluster D.ClusterName) (D.ASGName, []D.InstanceID, error) {
	url := s.activeASGURL(app, string(account), string(cluster), cloudProvider, string(region))

	resp, err := s.client.Get(url)
	if err != nil {
		return "", nil, errors.Wrapf(err, "http get failed at %s", url)
	}

	defer func() {
		if cerr := resp.Body.Close(); cerr != nil && err == nil {
			err = errors.Wrapf(err, "body close failed at %s", url)
		}
	}()

	if resp.StatusCode != http.StatusOK {
		return "", nil, errors.Errorf("unexpected response code (%d) from %s", resp.StatusCode, url)
	}

	body, err := ioutil.ReadAll(resp.Body)
	if err != nil {
		return "", nil, errors.Wrap(err, fmt.Sprintf("body read failed at %s", url))
	}

	var data struct {
		Name      string
		Instances []struct{ Name string }
	}

	err = json.Unmarshal(body, &data)

View on GitHub (pinned to eaa28fb761)

Solutions

Read the full body (ioutil.ReadAll) before the deferred close runs.
Check for concurrent use of the same http.Client/connection; share the client but not responses across goroutines.
Retry the request; body-close failures after keep-alive races are usually transient.
Upgrade the Go version / net/http usage if hitting known keep-alive close bugs.

Example fix

// before
defer func() {
    if cerr := resp.Body.Close(); cerr != nil && err == nil {
        err = errors.Wrapf(err, "body close failed at %s", url)
    }
}()
// after
defer func() {
    if cerr := resp.Body.Close(); cerr != nil && err == nil {
        err = errors.Wrapf(cerr, "body close failed at %s", url) // wrap cerr, not err
    }
}()
io.Copy(ioutil.Discard, resp.Body) // drain body before close

Defensive patterns

Strategy: try-catch

Try / catch

// Go: inspect the wrapped error; treat as transient
asg, ids, err := sp.GetInstanceIDs(app, acct, cp, region, cluster)
if err != nil {
    if strings.Contains(err.Error(), "body close failed") {
        // usually transient; log and retry once
        log.Printf("transient body-close failure, retrying: %v", err)
        return sp.GetInstanceIDs(app, acct, cp, region, cluster)
    }
    return err
}

Prevention

Always fully read or drain the response body before close
Share http.Client across goroutines but never responses
Keep Go's net/http up to date
Log full error chains to spot recurring close failures

When it happens

Trigger: Calling GetInstanceIDs when the response was received but resp.Body.Close() returns a non-nil error — typically after the connection was reused/aborted, or the body was never fully read before close, or a keep-alive connection was closed unexpectedly.

Common situations: Server closes keep-alive connections mid-flight; reading was skipped or incomplete before close; Go http client connection-reuse races under concurrency.

Related errors

AI-assisted analysis of Netflix/chaosmonkey@eaa28fb761 (2026-09-03). Data as JSON: /api/errors/66064afa74afd81e. Report an issue: GitHub.