Gabriel39 commented on code in PR #66348:
URL: https://github.com/apache/doris/pull/66348#discussion_r3703960180
##########
be/src/exec/sink/writer/async_result_writer.cpp:
##########
@@ -54,9 +56,12 @@ Status AsyncResultWriter::sink(Block* block, bool eos) {
if (_is_finished()) {
_dependency->set_ready();
}
- if (rows) {
- _memory_used_counter->update(add_block->allocated_bytes());
- _data_queue.emplace_back(std::move(add_block));
+ if (rows || eos) {
+ if (rows) {
+ _memory_used_counter->update(add_block->allocated_bytes());
+ }
+ _data_queue.emplace_back(QueuedBlock {
Review Comment:
Fixed in af29c53bf8. Every terminal writer exit now drains the queue under
the writer lock, decrements queued-block accounting, destroys reservation
tokens immediately, resets admission state, and wakes the dependency. The
regression uses RAII reservation tokens to verify that an abandoned queue
returns to baseline.
##########
be/src/io/fs/azure_obj_storage_client.cpp:
##########
@@ -216,35 +237,78 @@ ObjectStorageResponse
AzureObjStorageClient::put_object(const ObjectStoragePathO
ObjectStorageUploadResponse AzureObjStorageClient::upload_part(const
ObjectStoragePathOptions& opts,
std::string_view stream,
int part_num) {
- auto client = _client->GetBlockBlobClient(opts.key);
+ DCHECK(opts.upload_id.has_value());
+ auto client =
_client->GetBlockBlobClient(azure_multipart_temp_key(opts.key,
*opts.upload_id));
+ std::string block_id = azure_block_id(opts, part_num);
auto resp = do_azure_client_call(
[&]() {
Azure::Core::IO::MemoryBodyStream memory_body(
reinterpret_cast<const uint8_t*>(stream.data()),
stream.size());
// The blockId must be base64 encoded
SCOPED_BVAR_LATENCY(s3_bvar::s3_multi_part_upload_latency);
- client.StageBlock(base64_encode_part_num(part_num),
memory_body);
+ client.StageBlock(block_id, memory_body);
+ if (opts.deferred_completion) {
Review Comment:
Fixed in af29c53bf8. The protocol is now gated by an explicit coordinator
capability: a new BE rejects Azure Hive writes before file creation when the FE
cannot publish exact staged block IDs. New coordinators carry exact IDs, and an
expired lease is reacquired with the same proposed ID only when no competing
writer owns the target; otherwise completion fails closed. FE and BE
regressions cover capability rejection and expired-lease handling.
##########
fe/fe-connector/fe-connector-iceberg/src/main/java/org/apache/doris/connector/iceberg/IcebergConnectorTransaction.java:
##########
@@ -304,10 +304,49 @@ private void applyBeginGuards(IcebergWriteContext ctx,
String tableName) {
throw new IllegalArgumentException(branchName
+ " is a tag, not a branch. Tags cannot be targets
for producing snapshots");
}
+ this.baseSnapshotId = op == WriteOperation.OVERWRITE
+ ? resolveOverwriteBaseSnapshot(ctx,
branchRef.snapshotId(), tableName) : null;
} else {
this.branchName = null;
+ this.baseSnapshotId = op == WriteOperation.OVERWRITE
+ ? resolveOverwriteBaseSnapshot(ctx,
getSnapshotIdIfPresent(table), tableName) : null;
+ }
+ }
+ }
+
+ private Long resolveOverwriteBaseSnapshot(IcebergWriteContext ctx, Long
targetHead, String tableName) {
+ if (!ctx.isReadSnapshotPinned()) {
+ return targetHead;
+ }
+ long readSnapshotId = ctx.getReadSnapshotId();
+ if (readSnapshotId < 0) {
+ // An explicit empty read must conflict with any snapshot created
before beginWrite.
+ if (targetHead != null) {
+ throw new DorisConnectorException("Iceberg table " + tableName
Review Comment:
Fixed in af29c53bf8. The explicit empty generation remains pinned as -1
across transaction creation, so validateFromSnapshot keeps the OCC fence after
BaseTable.newTransaction refreshes. The injected race test commits the first
snapshot inside newTransaction and verifies that overwrite commit is rejected.
##########
be/src/exec/operator/spill_iceberg_table_sink_operator.cpp:
##########
@@ -55,47 +57,62 @@ size_t
SpillIcebergTableSinkLocalState::get_reserve_mem_size(RuntimeState* state
if (!_writer) {
return 0;
}
- auto current_writer = _writer->current_writer();
- auto* sort_writer =
dynamic_cast<VIcebergSortWriter*>(current_writer.get());
- if (!sort_writer) {
- return 0;
+ std::vector<IcebergSorterReserveMemory> per_partition_reservations;
+ auto active_writers = _writer->active_writers();
+ per_partition_reservations.reserve(active_writers->size());
+ for (const auto& writer : *active_writers) {
+ if (auto* sort_writer =
dynamic_cast<VIcebergSortWriter*>(writer.get())) {
+ auto reservation =
sort_writer->get_reserve_mem_size_components(state, eos);
+ per_partition_reservations.push_back(
+ {.retained_growth = reservation.retained_growth,
+ .transient_workspace = reservation.transient_workspace});
+ }
}
-
- return sort_writer->get_reserve_mem_size(state, eos);
+ // Column growth remains in every touched sorter, while sorting workspace
is reused by serial dispatch.
+ // The final queued item may contain rows and also owns the reservation
used by async finish().
Review Comment:
Fixed in af29c53bf8. EOS reservation now includes one spill buffer for every
stream in the configured merge fan-in plus the emitted block, including a
pending sorter run. The empty-sorter regression covers 12 spill runs with a 64
MiB fan-in limit and verifies a 72 MiB workspace reservation.
##########
fe/fe-connector/fe-connector-iceberg/src/main/java/org/apache/doris/connector/iceberg/action/IcebergRemoveOrphanFilesAction.java:
##########
@@ -0,0 +1,344 @@
+// Licensed to the Apache Software Foundation (ASF) under one
+// or more contributor license agreements. See the NOTICE file
+// distributed with this work for additional information
+// regarding copyright ownership. The ASF licenses this file
+// to you under the Apache License, Version 2.0 (the
+// "License"); you may not use this file except in compliance
+// with the License. You may obtain a copy of the License at
+//
+// http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing,
+// software distributed under the License is distributed on an
+// "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+// KIND, either express or implied. See the License for the
+// specific language governing permissions and limitations
+// under the License.
+
+package org.apache.doris.connector.iceberg.action;
+
+import org.apache.doris.connector.api.ConnectorColumn;
+import org.apache.doris.connector.api.ConnectorSession;
+import org.apache.doris.connector.api.ConnectorType;
+import org.apache.doris.connector.api.DorisConnectorException;
+import org.apache.doris.connector.api.pushdown.ConnectorPredicate;
+import org.apache.doris.foundation.util.ArgumentParsers;
+
+import com.google.common.collect.Lists;
+import org.apache.iceberg.DataFile;
+import org.apache.iceberg.DeleteFile;
+import org.apache.iceberg.ManifestContent;
+import org.apache.iceberg.ManifestFile;
+import org.apache.iceberg.ManifestFiles;
+import org.apache.iceberg.ManifestReader;
+import org.apache.iceberg.ReachableFileUtil;
+import org.apache.iceberg.Snapshot;
+import org.apache.iceberg.Table;
+import org.apache.iceberg.TableProperties;
+import org.apache.iceberg.io.FileInfo;
+import org.apache.iceberg.io.SupportsPrefixOperations;
+import org.apache.iceberg.util.PropertyUtil;
+
+import java.io.IOException;
+import java.net.URI;
+import java.time.Duration;
+import java.util.ArrayList;
+import java.util.HashSet;
+import java.util.LinkedHashMap;
+import java.util.List;
+import java.util.Locale;
+import java.util.Map;
+import java.util.Objects;
+import java.util.Set;
+
+/** Safely lists or deletes old files that are unreachable from every retained
snapshot. */
+public class IcebergRemoveOrphanFilesAction extends BaseIcebergAction {
+ private static final long MIN_RETENTION_MS =
Duration.ofHours(24).toMillis();
+ public static final String OLDER_THAN = "older_than";
+ public static final String LOCATION = "location";
+ public static final String DRY_RUN = "dry_run";
+
+ public IcebergRemoveOrphanFilesAction(Map<String, String> properties,
List<String> partitionNames,
+ ConnectorPredicate whereCondition) {
+ super("remove_orphan_files", properties, partitionNames,
whereCondition);
+ }
+
+ @Override
+ protected void registerIcebergArguments() {
+ namedArguments.registerRequiredArgument(OLDER_THAN, "Creation time
cutoff in milliseconds",
+ ArgumentParsers.nonNegativeLong(OLDER_THAN));
+ namedArguments.registerOptionalArgument(LOCATION, "Prefix within the
table location",
+ null, ArgumentParsers.nonEmptyString(LOCATION));
+ namedArguments.registerOptionalArgument(DRY_RUN, "Only count orphan
files", true,
+ ArgumentParsers.booleanValue(DRY_RUN));
+ }
+
+ @Override
+ protected void validateIcebergAction() {
+ validateNoPartitions();
+ validateNoWhereCondition();
+ String location = namedArguments.getString(LOCATION);
+ if (location != null) {
+ try {
+ normalizeLocation(location);
+ } catch (IllegalArgumentException e) {
+ throw new DorisConnectorException("Invalid location URI: " +
location, e);
+ }
+ }
+ }
+
+ @Override
+ protected List<String> executeAction(Table table, ConnectorSession
session) {
+ if (!(table.io() instanceof SupportsPrefixOperations)) {
+ throw new DorisConnectorException("remove_orphan_files requires
FileIO prefix listing support");
+ }
+ if (!PropertyUtil.propertyAsBoolean(table.properties(),
TableProperties.GC_ENABLED,
+ TableProperties.GC_ENABLED_DEFAULT)) {
+ // A GC-disabled table may share files with another table, so no
destructive scan is safe.
+ throw new DorisConnectorException("Cannot remove orphan files:
Iceberg GC is disabled");
+ }
+ List<String> scanLocations = resolveScanLocations(table);
+
+ try {
+ ReachableIndex reachable = new
ReachableIndex(collectReachableFiles(table));
+ long orphanCount = 0;
+ long deletedCount = 0;
+ long olderThan = namedArguments.getLong(OLDER_THAN);
+ // The SQL procedure needs a retention fence because concurrent
uploads are not reachable until commit.
+ if (olderThan > System.currentTimeMillis() - MIN_RETENTION_MS) {
+ throw new DorisConnectorException(
+ "older_than must retain at least 24 hours of files");
+ }
+ boolean dryRun = namedArguments.getBoolean(DRY_RUN);
+ Set<String> visitedFiles = new HashSet<>();
+ for (String scanLocation : scanLocations) {
+ // Object stores use raw prefix matching, so the separator
excludes sibling prefixes.
+ String listingPrefix = scanLocation.endsWith("/") ?
scanLocation : scanLocation + "/";
+ for (FileInfo file : ((SupportsPrefixOperations)
table.io()).listPrefix(listingPrefix)) {
+ if (visitedFiles.add(file.location()) &&
file.createdAtMillis() < olderThan
+ && !isReachable(file.location(), reachable)) {
+ orphanCount++;
+ if (!dryRun) {
+ table.io().deleteFile(file.location());
+ deletedCount++;
+ }
+ }
+ }
+ }
+ return Lists.newArrayList(String.valueOf(orphanCount),
String.valueOf(deletedCount));
+ } catch (Exception e) {
+ throw new DorisConnectorException("Failed to remove orphan files:
" + e.getMessage(), e);
+ }
+ }
+
+ private List<String> resolveScanLocations(Table table) {
+ List<String> ownedRoots = resolveOwnedRoots(table.location(),
table.properties());
+
+ String requested = namedArguments.getString(LOCATION);
+ if (requested != null) {
+ String normalized = normalizeLocation(requested);
+ boolean owned = ownedRoots.stream().anyMatch(root ->
isWithin(normalized, root));
+ if (!owned) {
+ throw new DorisConnectorException(
+ "location must be within an Iceberg table-owned
metadata or data location");
+ }
+ return Lists.newArrayList(normalized);
+ }
+
+ return minimalOwnedRoots(ownedRoots);
+ }
+
+ static List<String> resolveOwnedRoots(String tableLocation, Map<String,
String> properties) {
+ String tableRoot = normalizeLocation(tableLocation);
+ String dataRoot = normalizeLocation(resolveDataLocation(properties,
tableRoot));
Review Comment:
Fixed in af29c53bf8. Inferred shared object-store roots now fail closed
because a table suffix cannot prove ownership. Cleanup outside the table root
requires an explicit location plus allow_unsafe_location=true. The regression
creates two tables with the same db/table suffix under different catalogs and
verifies that neither file is deleted.
##########
be/src/runtime/runtime_state.cpp:
##########
@@ -52,12 +52,37 @@
#include "runtime/thread_context.h"
#include "storage/id_manager.h"
#include "storage/storage_engine.h"
+#include "util/thrift_util.h"
#include "util/timezone_utils.h"
#include "util/uid_util.h"
namespace doris {
using namespace ErrorCode;
+Status RuntimeState::add_iceberg_commit_datas(TIcebergCommitData
iceberg_commit_data) {
+ ThriftSerializer serializer(false, 256);
+ uint32_t serialized_size = 0;
+ uint8_t* buffer = nullptr;
+ RETURN_IF_ERROR(serializer.serialize(&iceberg_commit_data,
&serialized_size, &buffer));
+
+ constexpr size_t report_envelope_headroom = 1024 * 1024;
+ const size_t thrift_limit =
static_cast<size_t>(std::max(config::thrift_max_message_size, 0));
Review Comment:
Fixed in af29c53bf8. The complete TReportExecStatusParams is serialized
immediately before RPC and checked against the smaller coordinator/BE Thrift
limit, including a conservative RPC envelope. Closed files remain
rollback-owned until the final successful coordinator acknowledgement; final
size, submission, connection, RPC, and error-report failures execute cleanup
exactly once. The regression includes a 2 MiB non-Iceberg error log and shared
cleanup ownership across task states.
--
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]
