dolysis
/
rocksdb
mirror of https://github.com/facebook/rocksdb.git
3
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

Enable reopen with un-synced data loss in crash test (#12746) 

Summary:
**Context/Summary:**
https://github.com/facebook/rocksdb/pull/12567 disabled reopen with un-synced data loss in crash test since we discovered un-synced WAL loss and we currently don't support prefix recovery in reopen. This PR explicitly sync WAL data before close to avoid such data loss case from happening and add back the testing coverage.

Pull Request resolved: https://github.com/facebook/rocksdb/pull/12746

Test Plan: CI

Reviewed By: ajkr

Differential Revision: D58326890

Pulled By: hx235

fbshipit-source-id: 0865f715e97c5948d7cb3aea62fe2a626cb6522a
This commit is contained in:
Hui Xiao 2024-06-10 12:35:53 -07:00 committed by Facebook GitHub Bot
parent 43906597f5
commit d3c4b7fe0b
2 changed files with 21 additions and 5 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

21
db_stress_tool/db_stress_test_base.cc
View File

@ -3398,6 +3398,27 @@ void StressTest::Reopen(ThreadState* thread) {
} }
column_families_.clear(); column_families_.clear();
// Currently reopen does not restore expected state
// with potential data loss in mind like the first open before
// crash-recovery verification does. Therefore it always expects no data loss
// and we should ensure no data loss in testing.
// TODO(hx235): eliminate the FlushWAL(true /* sync */)/SyncWAL() below
if (!FLAGS_disable_wal) {
Status s;
if (FLAGS_manual_wal_flush_one_in > 0) {
s = db_->FlushWAL(/*sync=*/true);
} else {
s = db_->SyncWAL();
}
if (!s.ok()) {
fprintf(stderr,
"Error persisting WAL data which is needed before reopening the "
"DB: %s\n",
s.ToString().c_str());
exit(1);
}
}
if (thread->rand.OneIn(2)) { if (thread->rand.OneIn(2)) {
Status s = db_->Close(); Status s = db_->Close();
if (!s.ok()) { if (!s.ok()) {

5
tools/db_crashtest.py
View File

@ -738,11 +738,6 @@ def finalize_and_sanitize(src_params):
# files, which would be problematic when unsynced data can be lost in # files, which would be problematic when unsynced data can be lost in
# crash recoveries. # crash recoveries.
dest_params["enable_compaction_filter"] = 0 dest_params["enable_compaction_filter"] = 0
# TODO(hx235): re-enable "reopen" after supporting unsynced data loss
# verification upon reopen. Currently reopen does not restore expected state
# with potential data loss in mind like start of each `./db_stress` run.
# Therefore it always expects no data loss.
dest_params["reopen"] = 0
# Only under WritePrepared txns, unordered_write would provide the same guarnatees as vanilla rocksdb # Only under WritePrepared txns, unordered_write would provide the same guarnatees as vanilla rocksdb
# unordered_write is only enabled with --txn, and txn_params disables inplace_update_support, so # unordered_write is only enabled with --txn, and txn_params disables inplace_update_support, so
# setting allow_concurrent_memtable_write=1 won't conflcit with inplace_update_support. # setting allow_concurrent_memtable_write=1 won't conflcit with inplace_update_support.