ttljob: fix frequency of logging progress

The progress is supposed to be logged periodically, but the logic to
check when the last log occurred was broken. This fixed it by resetting
the count of spans processed since the last log.

Release note: None

pkg/sql/ttl/ttljob/ttljob_processor.go

@@ -167,6 +167,7 @@ func (t *ttlProcessor) work(ctx context.Context) error {
 	}
 	var rowsDeletedSoFar atomic.Int64
 	var spansProccessedSoFar atomic.Int64
+	var spansProccessedSinceLastLog atomic.Int64
 
 	// Log progress approximately every 1% of spans processed.
 	updateEvery := max(1, processorSpanCount/100)
@@ -221,6 +222,7 @@ func (t *ttlProcessor) work(ctx context.Context) error {
 					// add before returning err in case of partial success
 					rowsDeletedSoFar.Add(spanRowCount)
 					spansProccessedSoFar.Add(1)
+					spansProccessedSinceLastLog.Add(1)
 					if err != nil {
 						// Continue until channel is fully read.
 						// Otherwise, the keys input will be blocked.
@@ -257,8 +259,10 @@ func (t *ttlProcessor) work(ctx context.Context) error {
 				// If the span has no rows, we still need to increment the processed
 				// count.
 				spansProccessedSoFar.Add(1)
+				spansProccessedSinceLastLog.Add(1)
 			}
-			if spansProccessedSoFar.Load() >= updateEvery {
+			if spansProccessedSinceLastLog.Load() >= updateEvery {
+				spansProccessedSinceLastLog.Store(0)
 				if err := logProgress(); err != nil {
 					return err
 				}