update comment on identifying data exception

sdks/java/io/google-cloud-platform/src/main/java/org/apache/beam/sdk/io/gcp/bigtable/BigtableIO.java

@@ -1332,12 +1332,6 @@ private BiConsumer<MutateRowResponse, Throwable> handleMutationException(
  return (MutateRowResponse result, Throwable exception) -> {
  if (exception != null) {
  if (isDataException(exception)) {
- // This case, of being an NotFoundException and not retryable,
- // indicates an issue with the data. In general, the likely cause is the user trying to
- // write to a column family that doesn't exist. The frequency of this is dependent on
- // how column families are determined upstream of the io. We can't know if this record
- // failed, or if it was another record in the batch. Thus, we retry each record
- // individually.
  retryIndividualRecord(record, window);
  } else {
  failures.add(new BigtableWriteException(record, exception));
@@ -1360,6 +1354,13 @@ private void retryIndividualRecord(
  }
  }
 
+ // This method checks if an exception is the result of an error in the data.
+ // We first check if the exception is retryable, because if it is, we want to retry it via the
+ // runner. If the method is retryable, we check if it is a NotFoundException, or if it's an
+ // InvalidArgumentException. A NotFoundException likely means that the mutation is trying to
+ // write to a column family that doesn't exist. An InvalidArgumentException means that the
+ // mutation itself is invalid, with either an empty row key, invalid timestamp (ts <= -2),
+ // an empty mutation, or a column qualifier that is too large.
  private static boolean isDataException(Throwable e) {
  if (e instanceof ApiException && !((ApiException) e).isRetryable()) {
  return e instanceof NotFoundException || e instanceof InvalidArgumentException;