weaviate/weaviate · error
read meta response body
Error message
read meta response body
What it means
This wraps io.ReadAll failures when consuming the /meta response body from the inference service. The HTTP exchange succeeded but the response stream broke mid-read — typically the remote side closed the connection abruptly or a proxy terminated it. The body is required because meta data is parsed as JSON from it.
Source
Thrown at modules/sum-transformers/client/sum_meta.go:37
"github.com/pkg/errors"
)
func (c *client) MetaInfo() (map[string]interface{}, error) {
req, err := http.NewRequestWithContext(context.Background(), "GET", c.url("/meta"), nil)
if err != nil {
return nil, errors.Wrap(err, "create GET meta request")
}
res, err := c.httpClient.Do(req)
if err != nil {
return nil, errors.Wrap(err, "send GET meta request")
}
defer res.Body.Close()
bodyBytes, err := io.ReadAll(res.Body)
if err != nil {
return nil, errors.Wrap(err, "read meta response body")
}
var resBody map[string]interface{}
if err := json.Unmarshal(bodyBytes, &resBody); err != nil {
return nil, errors.Wrap(err, "unmarshal meta response body")
}
return resBody, nil
}
View on GitHub (pinned to 75aa4b6d11)
Solutions
- Check inference container logs for panics or restarts during the request
- Bypass proxies and test /meta directly from the Weaviate container
- Retrain/redeploy the inference service if it is crash-looping
- Increase proxy/ingress timeouts between the two services
Defensive patterns
Strategy: retry
Try / catch
meta, err := mod.MetaInfo()
if err != nil {
if strings.Contains(err.Error(), "read meta response body") {
// broken response stream: retry the request
return retryWithBackoff(metaCall, 3)
}
return err
} Prevention
- Keep proxies/ingress between Weaviate and inference with adequate timeouts
- Watch inference container logs for mid-request crashes
- Retry idempotent GET /meta calls on IO errors
When it happens
Trigger: Inference container resets the connection while responding to /meta; intermediary (ingress, sidecar proxy) times out and cuts the body; extremely large/failed responses dropped.
Common situations: Unstable networks between Weaviate and the inference service; HTTP/1.0-style servers closing without content-length; ingress proxies with short idle timeouts.
Related errors
AI-assisted analysis of weaviate/weaviate@75aa4b6d11 (2026-09-04).
Data as JSON: /api/errors/9911567336317ede.
Report an issue: GitHub.