fe.log file frequently outputs Fail to publish version for tablets

[zxb2503]

Steps to reproduce the behavior (Required)

The leader fe fe.log file frequently outputs the following error message:

2024-10-23 18:06:24.778+08:00 ERROR (lake-publish-task-198|687) [PublishVersionDaemon.publishPartitionBatch():556] Fail to publish partition 5054238 of txnIds [117796]:
com.starrocks.rpc.RpcException: Fail to publish version for tablets [5055000]: prepare_primary_index: load primary index failed: Corruption: Bad segment file staros://5055000/data/000000000001525e_b0631f8c-61ff-4fcd-b895-b9faaf9fed79.dat: file size 0 < 12
be/src/storage/rowset/segment.cpp:243 Segment::parse_segment_footer(read_file.get(), &footer, footer_length_hint, partial_rowset_footer)
be/src/storage/lake/tablet_manager.cpp:719 segment->open(footer_size_hint, nullptr, lake_io_opts)
be/src/storage/lake/rowset.cpp:322 load_segments(&segments, false)
be/src/storage/persistent_index.cpp:3374 loader->rowset_iterator(pkey_schema, [&](const std::vector& itrs, uint32_t rowset_id) { for (size_t i = 0; i < itrs.size(); i++) { auto itr = itrs[i].get(); if (itr == nullptr) { continue; } while (true) { chunk->reset(); rowids.clear(); auto st = itr->get_next(chunk, &rowids); if (st.is_end_of_file()) { break; } else if (!st.ok()) { return st; } else { Column* pkc = nullptr; if (pk_column != nullptr) { pk_column->reset_column(); do { try { auto catched_setter_L3393 = CurrentThreadCatchSetter(true); { PrimaryKeyEncoder::encode(pkey_schema, chunk, 0, chunk->num_rows(), pk_column.get()); } } catch (std::bad_alloc const&) { MemTracker exceed_tracker = tls_exceed_mem_tracker; tls_exceed_mem_tracker = nullptr; tls_thread_status.set_is_catched(false); if (__builtin_expect(!!(exceed_tracker != nullptr), 1)) { return Status::MemoryLimitExceeded( exceed_tracker->err_msg(fmt::format("try consume:{}", tls_thread_status.try_consume_mem_size()))); } else { return Status::MemoryLimitExceeded("Mem usage has exceed the limit of BE"); } } catch (std::runtime_error const& e) { return Status::RuntimeError(fmt::format("Runtime error: {}", e.what())); } } while (0); pkc = pk_column.get(); } else { pkc = chunk->columns()[0].get(); } uint32_t rssid = rowset_id + i; uint64_t base = ((uint64_t)rssid) << 32; std::vector values; values.reserve(pkc->size()); while (false) static_cast(0), !((__builtin_expect(!(pkc->size() <= rowids.size()), 0))) ? (void)0 : google::logging::internal::LogMessageVoidify() & google::LogMessageFatal("be/src/storage/persistent_index.cpp", 3403).stream() << "Check failed: " "pkc->size() <= rowids.size()" " "; for (uint32_t i = 0; i < pkc->size(); i++) { values.emplace_back(base + rowids[i]); } Status st; if (pkc->is_binary()) { st = insert(pkc->size(), reinterpret_cast<const Slice*>(pkc->raw_data()), values.data(), false); } else { std::vector keys; do { try { auto catched_setter_L3412 = CurrentThreadCatchSetter(true); { keys.reserve(pkc->size()); } } catch (std::bad_alloc const&) { MemTracker* exceed_tracker = tls_exceed_mem_tracker; tls_exceed_mem_tracker = nullptr; tls_thread_status.set_is_catched(false); if (__builtin_expect(!!(exceed_tracker != nullptr), 1)) { return Status::MemoryLimitExceeded( exceed_tracker->err_msg(fmt::format("try consume:{}", tls_thread_status.try_consume_mem_size()))); } else { return Status::MemoryLimitExceeded("Mem usage has exceed the limit of BE"); } } catch (std::runtime_error const& e) { return Status::RuntimeError(fmt::format("Runtime error: {}", e.what())); } } while (0); const auto* fkeys = pkc->continuous_data(); for (size_t i = 0; i < pkc->size(); ++i) { keys.emplace_back(fkeys, _key_size); fkeys += _key_size; } st = insert(pkc->size(), reinterpret_cast<const Slice*>(keys.data()), values.data(), false); } if (!st.ok()) { google::LogMessage("be/src/storage/persistent_index.cpp", 3421, google::GLOG_ERROR).stream() << "load index failed: tablet=" << loader->tablet_id() << " rowset:" << rowset_id << " segment:" << i << " reason: " << st.to_string() << " current_size:" << size(); return st; } } } itr->close(); } return Status::OK(); })
be/src/storage/persistent_index.cpp:5273 _insert_rowsets(loader, pkey_schema, std::move(pk_column)), host: xx.xxx.xx.19
at com.starrocks.lake.Utils.publishVersionBatch(Utils.java:153) ~[starrocks-fe.jar:?]
at com.starrocks.transaction.PublishVersionDaemon.publishPartitionBatch(PublishVersionDaemon.java:546) ~[starrocks-fe.jar:?]
at com.starrocks.transaction.PublishVersionDaemon.lambda$publishLakeTransactionBatchAsync$12(PublishVersionDaemon.java:665) ~[starrocks-fe.jar:?]
at java.util.concurrent.CompletableFuture$AsyncSupply.run(CompletableFuture.java:1700) ~[?:?]
at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1128) ~[?:?]
at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:628) ~[?:?]
at java.lang.Thread.run(Thread.java:834) ~[?:?]

StarRocks version (Required)

StarRocks 3.3.4-56bcf6f run_mode = shared_data