Github user joshelser commented on a diff in the pull request:
https://github.com/apache/accumulo/pull/260#discussion_r119119845
--- Diff:
server/tserver/src/main/java/org/apache/accumulo/tserver/tablet/Tablet.java ---
@@ -715,47 +722,54 @@ Batch nextBatch(SortedKeyValueIterator<Key,Value>
iter, Range range, int num, Se
long maxResultsSize =
tableConfiguration.getAsBytes(Property.TABLE_SCAN_MAXMEM);
- if (columns.size() == 0) {
- iter.seek(range, LocalityGroupUtil.EMPTY_CF_SET, false);
- } else {
- iter.seek(range, LocalityGroupUtil.families(columns), true);
- }
-
Key continueKey = null;
boolean skipContinueKey = false;
boolean endOfTabletReached = false;
- while (iter.hasTop()) {
- value = iter.getTopValue();
- key = iter.getTopKey();
+ try {
+ if (columns.size() == 0) {
+ iter.seek(range, LocalityGroupUtil.EMPTY_CF_SET, false);
--- End diff --
If `seek` throws the `ScanYieldException` and there exists a Key in the
data which is the same as the startKey of the Range, I believe we would miss
data.
`skipContinueKey` is set to true (in the catch below), but when we're
`seek()`'ing here, we haven't yet consumed that top key. Thus, on a re-seek, we
would skip past it.
My hunch is that you need a two catch statements: one here that does not
skip the continueKey and the one below around the `while(hasTop)` (as-is).
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at infrastruct...@apache.org or file a JIRA ticket
with INFRA.
---
