Github user aweisberg commented on a diff in the pull request:
https://github.com/apache/cassandra/pull/269#discussion_r218561162
--- Diff: src/java/org/apache/cassandra/dht/RangeStreamer.java ---
@@ -600,39 +628,38 @@ public StreamResultFuture fetchAsync()
sources.asMap().forEach((source, fetchReplicas) -> {
// filter out already streamed ranges
- RangesAtEndpoint available =
stateStore.getAvailableRanges(keyspace,
StorageService.instance.getTokenMetadata().partitioner);
+ Pair<Set<Range<Token>>, Set<Range<Token>>> available =
stateStore.getAvailableRanges(keyspace, metadata.partitioner);
Predicate<FetchReplica> isAvailable = fetch -> {
- Replica availableRange =
available.byRange().get(fetch.local.range());
- if (availableRange == null)
+ boolean isInFull =
available.left.contains(fetch.local.range());
+ boolean isInTrans =
available.right.contains(fetch.local.range());
+
+ if (!isInFull && !isInTrans)
//Range is unavailable
return false;
+
+ assert isInFull != isInTrans : "Range can't be
simultaneously full and transient: " + isInFull + " " + isInTrans;
--- End diff --
It can right? Isn't that the point of this whole exercise? We need to fetch
it twice and it might have already been fetched both fully and transiently.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
