github-actions[bot] commented on code in PR #69134:
URL: https://github.com/apache/doris/pull/69134#discussion_r4292420693
##########
be/src/cloud/cloud_schema_change_job.cpp:
##########
@@ -54,12 +54,30 @@ static constexpr int SCHEMA_CHANGE_DELETE_BITMAP_LOCK_ID =
-2;
std::unique_ptr<SchemaChange> get_sc_procedure(const BlockChanger& changer,
bool sc_sorting,
int64_t mem_limit) {
if (sc_sorting) {
- return std::make_unique<VBaseSchemaChangeWithSorting>(changer,
mem_limit);
+ return std::make_unique<VCloudSchemaChangeWithSorting>(changer,
mem_limit);
}
// else sc_directly
return std::make_unique<VSchemaChangeDirectly>(changer);
}
+Status VCloudSchemaChangeWithSorting::_inner_process(RowsetReaderSharedPtr
rowset_reader,
+ RowsetWriter*
rowset_writer,
+ BaseTabletSPtr new_tablet,
+ TabletSchemaSPtr
base_tablet_schema,
+ TabletSchemaSPtr
new_tablet_schema) {
+ Defer defer {[&]() {
+ // remove the intermediate rowsets generated by internal sorting
+ for (auto& rowset : _src_rowsets) {
Review Comment:
[P2] Include failed spill index files in cleanup
When internal sorting finishes an indexed segment and then fails while
opening/writing the next one (for example, ENOSPC during a cloud base-table key
reorder with a retained inverted index), `_internal_sorting()` returns before
that spill is pushed into `_src_rowsets`. The failed writer's existing
destructor removes only `.dat` files; completed `.idx` outputs are not unlinked
by their writers. This cleanup therefore still leaves those index files
occupying the cache disk until BE restart. Please cover index outputs from
failed spill writers too and add an indexed failure before `build()`; the new
failure point only runs after all spills have built.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
