Details
-
Bug
-
Resolution: Unresolved
-
Major
-
7.2.4
-
None
-
Untriaged
-
0
-
Unknown
Description
- If scan for partitioned index times out on few of the partitions during scanPipeline.Execute(), error is not logged on indexer side.
- Only ItemWriter.Shutdown is done in CancelCallback which sends err to IndexConnection.
- Increment in num_scan_timeouts can be seen in stats.
func (s *scanCoordinator) handleScanRequest(req *ScanRequest, w ScanResponseWriter,
|
is IndexSnapshot, t0 time.Time) {
|
waitTime := time.Now().Sub(t0)
|
scanPipeline := NewScanPipeline(req, w, is, s.config.Load())
|
cancelCb := NewCancelCallback(req, func(e error) {
|
scanPipeline.Cancel(e)
|
})
|
cancelCb.Run()
|
defer cancelCb.Done()
|
|
err := scanPipeline.Execute()
|
...
|
if err != nil { |
status := fmt.Sprintf("(error = %s)", err) |
logging.LazyVerbose(func() string {
|
return fmt.Sprintf("%s RESPONSE rows:%d, scanned:%d, waitTime:%v, totalTime:%v, status:%s, requestId:%s", |
req.LogPrefix, scanPipeline.RowsReturned(), scanPipeline.RowsScanned(), waitTime, scanTime, status, req.RequestId)
|
})
|
s.updateErrStats(req, err)
|
}
|
Attachments
Gerrit Reviews
For Gerrit Dashboard: MB-61600 | ||||||
---|---|---|---|---|---|---|
# | Subject | Branch | Project | Status | CR | V |
209010,4 | MB-61600: Log error if index scan timing out during scanpipeline.Execution | unstable | indexing | Status: NEW | +2 | 0 |