github-actions[bot] commented on code in PR #66348:
URL: https://github.com/apache/doris/pull/66348#discussion_r3703191085
##########
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:
Follow-up: the new-BE/old-FE path only fails closed while this 60-second
lease is active. A file can finish more than a minute before the
transaction-wide FE commit, with no renewal in between. After expiry, the old
FE's unfenced `commitBlockList` is accepted; on an existing legacy target (or
one with residual legacy blocks), its deterministic part IDs select the old
blocks rather than the new BE's namespaced blocks, so Hive can commit metadata
while stale bytes are republished. Please gate this protocol on an
exact-ID-aware coordinator or keep a renewable fence alive until that
coordinator completes it, and test the expired-lease rolling path.
##########
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:
Follow-up: matching the exact `parent/table` context still does not prove
exclusivity. Iceberg derives that context from only the last two table-location
components, so tables such as `.../catalog-a/db/t` and `.../catalog-b/db/t` can
share one `write.data.path` and generate the same `<hash>/db/t/...` pattern.
Running this action for either table then treats the other's old live files as
owned-but-unreachable and can delete them. Please fail closed for a shared
external root unless context ownership is established outside this suffix (or
require the explicit unsafe-location path), and add a same-suffix two-table
test.
##########
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:
Follow-up: the receiver-limit propagation is fixed, but the constant 1 MiB
allowance is not a bound on the rest of `TReportExecStatusParams`. The callback
adds variable-length fields after this guard, including `error_log` (UDF
warnings are arbitrary strings), first-error/load reports, export files, and
other commit vectors. Commit metadata can therefore fill the admitted budget
and still make the final RPC exceed the receiver limit after successful writer
close has released file-cleanup ownership, stranding the files. Please budget
the complete binary report (or retain cleanup ownership until acknowledgement)
and add a near-limit test with a >1 MiB non-Iceberg envelope.
##########
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:
[P2] Release queued reservation tokens when asynchronous `open()` fails. The
producer can enqueue here while the writer thread is still opening;
`force_close()` then makes `while (_writer_status.ok())` skip the only
`_data_queue.clear()` path, and the token stays charged until writer
destruction even though the finish dependency is ready. This can retain a large
query/process/workload-group reservation throughout cancellation. Please clear
or swap the queue on every terminal exit and add an enqueue-before-open-failure
test that verifies the accounting returns to baseline.
##########
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:
[P1] Reserve the final spill-merge fan-in on EOS. When a sorter has already
spilled its in-memory block, its reserve components are zero, so this path
retains only the minimum allowance. `finish()` nevertheless enters
`_combine_files_output()`, whose merger constructs every input cursor and
immediately reads one spill block per run (up to the configured merge fan-in,
64 MiB by default) before producing output. That work runs under this
transferred reservation but is materially under-reserved and can cross a hard
query/process limit. Please include EOS merge inputs/output workspace in the
estimate and cover an empty sorter with several spill runs under a tight limit.
##########
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:
[P1] Preserve the explicit-empty generation fence across the transaction
refresh. `applyBeginGuards()` observes an empty head and converts the pinned
`-1` to `null` here, but `openTransaction()` immediately follows and
`newTransaction()` refreshes the mutable table. If another writer creates the
first snapshot in that interval, it becomes this transaction's base; the
overwrite paths then skip `validateFromSnapshot` because `baseSnapshotId` is
null and can replace that concurrent data. Please retain the `-1` OCC sentinel
(as the RowDelta path does) or perform the empty-head check against the
refreshed transaction base, with an injected check-to-refresh race test.
--
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]
