dolysis/rocksdb
3
0
Fork 0
mirror of https://github.com/facebook/rocksdb.git synced 2024-11-26 07:30:54 +00:00
Code Issues Packages Projects Releases Wiki Activity

db_stress tolerate incomplete tail records in trace file (#9316)

Browse source 
Summary:
I saw the following error when running crash test for a while with
unsynced data loss:

```
Error restoring historical expected values: Corruption: Corrupted trace file.
```

The trace file turned out to have an incomplete tail record. This is
normal considering blackbox kills `db_stress` while trace can be
ongoing.

In the case where the trace file is not otherwise corrupted, there
should be enough records already seen to sync up the expected state with
the recovered DB. This PR ignores any `Status::Corruption` the
`Replayer` returns when that happens.

Pull Request resolved: https://github.com/facebook/rocksdb/pull/9316

Reviewed By: jay-zhuang

Differential Revision: D33230579

Pulled By: ajkr

fbshipit-source-id: 9814af4e39e57f00d85be7404363211762f9b41b
This commit is contained in:
Andrew Kryczka 2021-12-20 13:07:35 -08:00 committed by Facebook GitHub Bot
parent 791723c1ec
commit b448b71222
1 changed files with 13 additions and 5 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

18
db_stress_tool/expected_state.cc
View file

@ -338,13 +338,14 @@ class ExpectedStateTraceRecordHandler : public TraceRecord::Handler,
ExpectedStateTraceRecordHandler(uint64_t max_write_ops, ExpectedState* state)
: max_write_ops_(max_write_ops), state_(state) {}
~ExpectedStateTraceRecordHandler() {
assert(num_write_ops_ == max_write_ops_);
}
~ExpectedStateTraceRecordHandler() { assert(IsDone()); }
// True if we have already reached the limit on write operations to apply.
bool IsDone() { return num_write_ops_ == max_write_ops_; }
Status Handle(const WriteQueryTraceRecord& record,
std::unique_ptr<TraceRecordResult>* /* result */) override {
if (num_write_ops_ == max_write_ops_) {
if (IsDone()) {
return Status::OK();
}
WriteBatch batch(record.GetWriteBatchRep().ToString());
@ -466,7 +467,7 @@ Status FileExpectedStateManager::Restore(DB* db) {
{
std::unique_ptr<Replayer> replayer;
std::unique_ptr<ExpectedState> state;
std::unique_ptr<TraceRecord::Handler> handler;
std::unique_ptr<ExpectedStateTraceRecordHandler> handler;
if (s.ok()) {
state.reset(new FileExpectedState(latest_file_temp_path, max_key_,
num_column_families_));
@ -494,6 +495,13 @@ Status FileExpectedStateManager::Restore(DB* db) {
std::unique_ptr<TraceRecordResult> res;
record->Accept(handler.get(), &res);
}
if (s.IsCorruption() && handler->IsDone()) {
// There could be a corruption reading the tail record of the trace due to
// `db_stress` crashing while writing it. It shouldn't matter as long as
// we already found all the write ops we need to catch up the expected
// state.
s = Status::OK();
}
if (s.IsIncomplete()) {
// OK because `Status::Incomplete` is expected upon finishing all the
// trace records.