Tables with Inverted Index fails with BAD_SIZE_OF_FILE_IN_DATA_PART error when CH restarted or attach table

[zhangjmruc]

Describe the unexpected behaviour
I tried to create a table with Inverted Index, and insert some rows. But when CH restarts or attach table, there is no rows in this table.

How to reproduce
SET allow_experimental_inverted_index = true;

CREATE TABLE tab
(
key UInt64,
str String,
INDEX inv_idx(str) TYPE inverted(0) GRANULARITY 1
)
ENGINE = MergeTree
ORDER BY key;

INSERT INTO tab(key, str) values (1, 'Hello World');

detach table tab;
attach table tab;

select count() from tab;

• Which ClickHouse server version to use
  Master: 23.3.1.2537

Expected behavior
The table with Inverted Index can be attached correctly, or data can be loaded during CK start.

Error message and/or stacktrace

2023.03.21 13:58:05.017418 [ 920094 ] {f106f433-b9ed-43e1-a217-05ad7179c1e9} default.tab (a17bb23f-ff60-437b-bd9d-150eb366ed46): while loading part all_1_1_0 on path store/a17/a17bb23f-ff60-437b-bd9d-150eb366ed46/all_1_1_0: Code: 228. DB::Exception: store/a17/a17bb23f-ff60-437b-bd9d-150eb366ed46/all_1_1_0/skp_idx_inv_idx.gin_dict has unexpected size: 44 instead of 0. (BAD_SIZE_OF_FILE_IN_DATA_PART), Stack trace (when copying this message, always include the lines below):

0. Poco::Exception::Exception(String const&, int) @ 0x15871d5a in /home/zhang/clickhouse-test/clickhouse/bin/clickhouse
1. DB::Exception::Exception(DB::Exception::MessageMasked&&, int, bool) @ 0xe9605b5 in /home/zhang/clickhouse-test/clickhouse/bin/clickhouse
2. DB::Exception::Exception<std::__fs::filesystem::path, unsigned long&, unsigned long const&>(int, FormatStringHelperImpl<std::type_identitystd::__fs::filesystem::path::type, std::type_identity<unsigned long&>::type, std::type_identity<unsigned long const&>::type>, std::__fs::filesystem::path&&, unsigned long&, unsigned long const&) @ 0x1486f8b6 in /home/zhang/clickhouse-test/clickhouse/bin/clickhouse
3. DB::MergeTreeDataPartChecksum::checkSize(DB::IDataPartStorage const&, String const&) const @ 0x1486ade2 in /home/zhang/clickhouse-test/clickhouse/bin/clickhouse
4. DB::MergeTreeDataPartChecksums::checkSizes(DB::IDataPartStorage const&) const @ 0x1486b198 in /home/zhang/clickhouse-test/clickhouse/bin/clickhouse
5. DB::IMergeTreeDataPart::checkConsistencyBase() const @ 0x1477a927 in /home/zhang/clickhouse-test/clickhouse/bin/clickhouse
6. DB::MergeTreeDataPartCompact::checkConsistency(bool) const @ 0x1487199c in /home/zhang/clickhouse-test/clickhouse/bin/clickhouse
7. DB::IMergeTreeDataPart::loadColumnsChecksumsIndexes(bool, bool) @ 0x1476cfa2 in /home/zhang/clickhouse-test/clickhouse/bin/clickhouse
8. DB::MergeTreeData::loadDataPart(DB::MergeTreePartInfo const&, String const&, std::shared_ptrDB::IDisk const&, DB::MergeTreeDataPartState, std::mutex&) @ 0x147e411d in /home/zhang/clickhouse-test/clickhouse/bin/clickhouse
9. void std::__function::__policy_invoker<void ()>::__call_impl<std::__function::__default_alloc_func<DB::MergeTreeData::loadDataPartsFromDisk(ThreadPoolImpl<ThreadFromGlobalPoolImpl>&, unsigned long, std::queue<std::vector<std::shared_ptrDB::MergeTreeData::PartLoadingTree::Node, std::allocator<std::shared_ptrDB::MergeTreeData::PartLoadingTree::Node>>, std::deque<std::vector<std::shared_ptrDB::MergeTreeData::PartLoadingTree::Node, std::allocator<std::shared_ptrDB::MergeTreeData::PartLoadingTree::Node>>, std::allocator<std::vector<std::shared_ptrDB::MergeTreeData::PartLoadingTree::Node, std::allocator<std::shared_ptrDB::MergeTreeData::PartLoadingTree::Node>>>>>&, std::shared_ptr<DB::MergeTreeSettings const> const&)::$_19, void ()>>(std::__function::__policy_storage const*) @ 0x14849bd4 in /home/zhang/clickhouse-test/clickhouse/bin/clickhouse
10. ThreadPoolImpl<ThreadFromGlobalPoolImpl>::worker(std::__list_iterator<ThreadFromGlobalPoolImpl, void*>) @ 0xea1a4ed in /home/zhang/clickhouse-test/clickhouse/bin/clickhouse
11. ThreadFromGlobalPoolImpl::ThreadFromGlobalPoolImpl<void ThreadPoolImpl<ThreadFromGlobalPoolImpl>::scheduleImpl(std::function<void ()>, long, std::optional, bool)::'lambda0'()>(void&&)::'lambda'()::operator()() @ 0xea1d09b in /home/zhang/clickhouse-test/clickhouse/bin/clickhouse
12. ThreadPoolImplstd::thread::worker(std::__list_iterator<std::thread, void*>) @ 0xea183cd in /home/zhang/clickhouse-test/clickhouse/bin/clickhouse
13. void* std::__thread_proxy[abi:v15000]<std::tuple<std::unique_ptr<std::__thread_struct, std::default_deletestd::__thread_struct>, void ThreadPoolImplstd::thread::scheduleImpl(std::function<void ()>, long, std::optional, bool)::'lambda0'()>>(void*) @ 0xea1b6ae in /home/zhang/clickhouse-test/clickhouse/bin/clickhouse

[cangyin]

see #47607
if you encounter 'invalid segment id' error, see #47663

inverted index is still in experimental state.

[zhangjmruc]

@cangyin Thanks for your reply. I will close this issue.