rocksdb/file/filename.cc
Yanqin Jin a376c22066 Handle rename() failure in non-local FS (#8192)
Summary:
In a distributed environment, a file `rename()` operation can succeed on server (remote)
side, but the client can somehow return non-ok status to RocksDB. Possible reasons include
network partition, connection issue, etc. This happens in `rocksdb::SetCurrentFile()`, which
can be called in `LogAndApply() -> ProcessManifestWrites()` if RocksDB tries to switch to a
new MANIFEST. We currently always delete the new MANIFEST if an error occurs.

This is problematic in distributed world. If the server-side successfully updates the CURRENT
file via renaming, then a subsequent `DB::Open()` will try to look for the new MANIFEST and fail.

As a fix, we can track the execution result of IO operations on the new MANIFEST.
- If IO operations on the new MANIFEST fail, then we know the CURRENT must point to the original
  MANIFEST. Therefore, it is safe to remove the new MANIFEST.
- If IO operations on the new MANIFEST all succeed, but somehow we end up in the clean up
  code block, then we do not know whether CURRENT points to the new or old MANIFEST. (For local
  POSIX-compliant FS, it should still point to old MANIFEST, but it does not matter if we keep the
  new MANIFEST.) Therefore, we keep the new MANIFEST.
    - Any future `LogAndApply()` will switch to a new MANIFEST and update CURRENT.
    - If process reopens the db immediately after the failure, then the CURRENT file can point
      to either the new MANIFEST or the old one, both of which exist. Therefore, recovery can
      succeed and ignore the other.

Pull Request resolved: https://github.com/facebook/rocksdb/pull/8192

Test Plan: make check

Reviewed By: zhichao-cao

Differential Revision: D27804648

Pulled By: riversand963

fbshipit-source-id: 9c16f2a5ce41bc6aadf085e48449b19ede8423e4
2021-04-19 18:11:13 -07:00

478 lines
14 KiB
C++

// Copyright (c) 2011-present, Facebook, Inc. All rights reserved.
// This source code is licensed under both the GPLv2 (found in the
// COPYING file in the root directory) and Apache 2.0 License
// (found in the LICENSE.Apache file in the root directory).
//
// Copyright (c) 2011 The LevelDB Authors. All rights reserved.
// Use of this source code is governed by a BSD-style license that can be
// found in the LICENSE file. See the AUTHORS file for names of contributors.
#include "file/filename.h"
#include <cinttypes>
#include <ctype.h>
#include <stdio.h>
#include <vector>
#include "file/writable_file_writer.h"
#include "rocksdb/env.h"
#include "test_util/sync_point.h"
#include "util/stop_watch.h"
#include "util/string_util.h"
namespace ROCKSDB_NAMESPACE {
static const std::string kRocksDbTFileExt = "sst";
static const std::string kLevelDbTFileExt = "ldb";
static const std::string kRocksDBBlobFileExt = "blob";
// Given a path, flatten the path name by replacing all chars not in
// {[0-9,a-z,A-Z,-,_,.]} with _. And append '_LOG\0' at the end.
// Return the number of chars stored in dest not including the trailing '\0'.
static size_t GetInfoLogPrefix(const std::string& path, char* dest, int len) {
const char suffix[] = "_LOG";
size_t write_idx = 0;
size_t i = 0;
size_t src_len = path.size();
while (i < src_len && write_idx < len - sizeof(suffix)) {
if ((path[i] >= 'a' && path[i] <= 'z') ||
(path[i] >= '0' && path[i] <= '9') ||
(path[i] >= 'A' && path[i] <= 'Z') ||
path[i] == '-' ||
path[i] == '.' ||
path[i] == '_'){
dest[write_idx++] = path[i];
} else {
if (i > 0) {
dest[write_idx++] = '_';
}
}
i++;
}
assert(sizeof(suffix) <= len - write_idx);
// "\0" is automatically added by snprintf
snprintf(dest + write_idx, len - write_idx, suffix);
write_idx += sizeof(suffix) - 1;
return write_idx;
}
static std::string MakeFileName(uint64_t number, const char* suffix) {
char buf[100];
snprintf(buf, sizeof(buf), "%06llu.%s",
static_cast<unsigned long long>(number), suffix);
return buf;
}
static std::string MakeFileName(const std::string& name, uint64_t number,
const char* suffix) {
return name + "/" + MakeFileName(number, suffix);
}
std::string LogFileName(const std::string& name, uint64_t number) {
assert(number > 0);
return MakeFileName(name, number, "log");
}
std::string LogFileName(uint64_t number) {
assert(number > 0);
return MakeFileName(number, "log");
}
std::string BlobFileName(uint64_t number) {
assert(number > 0);
return MakeFileName(number, kRocksDBBlobFileExt.c_str());
}
std::string BlobFileName(const std::string& blobdirname, uint64_t number) {
assert(number > 0);
return MakeFileName(blobdirname, number, kRocksDBBlobFileExt.c_str());
}
std::string BlobFileName(const std::string& dbname, const std::string& blob_dir,
uint64_t number) {
assert(number > 0);
return MakeFileName(dbname + "/" + blob_dir, number,
kRocksDBBlobFileExt.c_str());
}
std::string ArchivalDirectory(const std::string& dir) {
return dir + "/" + ARCHIVAL_DIR;
}
std::string ArchivedLogFileName(const std::string& name, uint64_t number) {
assert(number > 0);
return MakeFileName(name + "/" + ARCHIVAL_DIR, number, "log");
}
std::string MakeTableFileName(const std::string& path, uint64_t number) {
return MakeFileName(path, number, kRocksDbTFileExt.c_str());
}
std::string MakeTableFileName(uint64_t number) {
return MakeFileName(number, kRocksDbTFileExt.c_str());
}
std::string Rocks2LevelTableFileName(const std::string& fullname) {
assert(fullname.size() > kRocksDbTFileExt.size() + 1);
if (fullname.size() <= kRocksDbTFileExt.size() + 1) {
return "";
}
return fullname.substr(0, fullname.size() - kRocksDbTFileExt.size()) +
kLevelDbTFileExt;
}
uint64_t TableFileNameToNumber(const std::string& name) {
uint64_t number = 0;
uint64_t base = 1;
int pos = static_cast<int>(name.find_last_of('.'));
while (--pos >= 0 && name[pos] >= '0' && name[pos] <= '9') {
number += (name[pos] - '0') * base;
base *= 10;
}
return number;
}
std::string TableFileName(const std::vector<DbPath>& db_paths, uint64_t number,
uint32_t path_id) {
assert(number > 0);
std::string path;
if (path_id >= db_paths.size()) {
path = db_paths.back().path;
} else {
path = db_paths[path_id].path;
}
return MakeTableFileName(path, number);
}
void FormatFileNumber(uint64_t number, uint32_t path_id, char* out_buf,
size_t out_buf_size) {
if (path_id == 0) {
snprintf(out_buf, out_buf_size, "%" PRIu64, number);
} else {
snprintf(out_buf, out_buf_size, "%" PRIu64
"(path "
"%" PRIu32 ")",
number, path_id);
}
}
std::string DescriptorFileName(const std::string& dbname, uint64_t number) {
assert(number > 0);
char buf[100];
snprintf(buf, sizeof(buf), "/MANIFEST-%06llu",
static_cast<unsigned long long>(number));
return dbname + buf;
}
std::string CurrentFileName(const std::string& dbname) {
return dbname + "/CURRENT";
}
std::string LockFileName(const std::string& dbname) {
return dbname + "/LOCK";
}
std::string TempFileName(const std::string& dbname, uint64_t number) {
return MakeFileName(dbname, number, kTempFileNameSuffix.c_str());
}
InfoLogPrefix::InfoLogPrefix(bool has_log_dir,
const std::string& db_absolute_path) {
if (!has_log_dir) {
const char kInfoLogPrefix[] = "LOG";
// "\0" is automatically added to the end
snprintf(buf, sizeof(buf), kInfoLogPrefix);
prefix = Slice(buf, sizeof(kInfoLogPrefix) - 1);
} else {
size_t len =
GetInfoLogPrefix(NormalizePath(db_absolute_path), buf, sizeof(buf));
prefix = Slice(buf, len);
}
}
std::string InfoLogFileName(const std::string& dbname,
const std::string& db_path, const std::string& log_dir) {
if (log_dir.empty()) {
return dbname + "/LOG";
}
InfoLogPrefix info_log_prefix(true, db_path);
return log_dir + "/" + info_log_prefix.buf;
}
// Return the name of the old info log file for "dbname".
std::string OldInfoLogFileName(const std::string& dbname, uint64_t ts,
const std::string& db_path, const std::string& log_dir) {
char buf[50];
snprintf(buf, sizeof(buf), "%llu", static_cast<unsigned long long>(ts));
if (log_dir.empty()) {
return dbname + "/LOG.old." + buf;
}
InfoLogPrefix info_log_prefix(true, db_path);
return log_dir + "/" + info_log_prefix.buf + ".old." + buf;
}
std::string OptionsFileName(const std::string& dbname, uint64_t file_num) {
char buffer[256];
snprintf(buffer, sizeof(buffer), "%s%06" PRIu64,
kOptionsFileNamePrefix.c_str(), file_num);
return dbname + "/" + buffer;
}
std::string TempOptionsFileName(const std::string& dbname, uint64_t file_num) {
char buffer[256];
snprintf(buffer, sizeof(buffer), "%s%06" PRIu64 ".%s",
kOptionsFileNamePrefix.c_str(), file_num,
kTempFileNameSuffix.c_str());
return dbname + "/" + buffer;
}
std::string MetaDatabaseName(const std::string& dbname, uint64_t number) {
char buf[100];
snprintf(buf, sizeof(buf), "/METADB-%llu",
static_cast<unsigned long long>(number));
return dbname + buf;
}
std::string IdentityFileName(const std::string& dbname) {
return dbname + "/IDENTITY";
}
// Owned filenames have the form:
// dbname/IDENTITY
// dbname/CURRENT
// dbname/LOCK
// dbname/<info_log_name_prefix>
// dbname/<info_log_name_prefix>.old.[0-9]+
// dbname/MANIFEST-[0-9]+
// dbname/[0-9]+.(log|sst|blob)
// dbname/METADB-[0-9]+
// dbname/OPTIONS-[0-9]+
// dbname/OPTIONS-[0-9]+.dbtmp
// Disregards / at the beginning
bool ParseFileName(const std::string& fname,
uint64_t* number,
FileType* type,
WalFileType* log_type) {
return ParseFileName(fname, number, "", type, log_type);
}
bool ParseFileName(const std::string& fname, uint64_t* number,
const Slice& info_log_name_prefix, FileType* type,
WalFileType* log_type) {
Slice rest(fname);
if (fname.length() > 1 && fname[0] == '/') {
rest.remove_prefix(1);
}
if (rest == "IDENTITY") {
*number = 0;
*type = kIdentityFile;
} else if (rest == "CURRENT") {
*number = 0;
*type = kCurrentFile;
} else if (rest == "LOCK") {
*number = 0;
*type = kDBLockFile;
} else if (info_log_name_prefix.size() > 0 &&
rest.starts_with(info_log_name_prefix)) {
rest.remove_prefix(info_log_name_prefix.size());
if (rest == "" || rest == ".old") {
*number = 0;
*type = kInfoLogFile;
} else if (rest.starts_with(".old.")) {
uint64_t ts_suffix;
// sizeof also counts the trailing '\0'.
rest.remove_prefix(sizeof(".old.") - 1);
if (!ConsumeDecimalNumber(&rest, &ts_suffix)) {
return false;
}
*number = ts_suffix;
*type = kInfoLogFile;
}
} else if (rest.starts_with("MANIFEST-")) {
rest.remove_prefix(strlen("MANIFEST-"));
uint64_t num;
if (!ConsumeDecimalNumber(&rest, &num)) {
return false;
}
if (!rest.empty()) {
return false;
}
*type = kDescriptorFile;
*number = num;
} else if (rest.starts_with("METADB-")) {
rest.remove_prefix(strlen("METADB-"));
uint64_t num;
if (!ConsumeDecimalNumber(&rest, &num)) {
return false;
}
if (!rest.empty()) {
return false;
}
*type = kMetaDatabase;
*number = num;
} else if (rest.starts_with(kOptionsFileNamePrefix)) {
uint64_t ts_suffix;
bool is_temp_file = false;
rest.remove_prefix(kOptionsFileNamePrefix.size());
const std::string kTempFileNameSuffixWithDot =
std::string(".") + kTempFileNameSuffix;
if (rest.ends_with(kTempFileNameSuffixWithDot)) {
rest.remove_suffix(kTempFileNameSuffixWithDot.size());
is_temp_file = true;
}
if (!ConsumeDecimalNumber(&rest, &ts_suffix)) {
return false;
}
*number = ts_suffix;
*type = is_temp_file ? kTempFile : kOptionsFile;
} else {
// Avoid strtoull() to keep filename format independent of the
// current locale
bool archive_dir_found = false;
if (rest.starts_with(ARCHIVAL_DIR)) {
if (rest.size() <= ARCHIVAL_DIR.size()) {
return false;
}
rest.remove_prefix(ARCHIVAL_DIR.size() + 1); // Add 1 to remove / also
if (log_type) {
*log_type = kArchivedLogFile;
}
archive_dir_found = true;
}
uint64_t num;
if (!ConsumeDecimalNumber(&rest, &num)) {
return false;
}
if (rest.size() <= 1 || rest[0] != '.') {
return false;
}
rest.remove_prefix(1);
Slice suffix = rest;
if (suffix == Slice("log")) {
*type = kWalFile;
if (log_type && !archive_dir_found) {
*log_type = kAliveLogFile;
}
} else if (archive_dir_found) {
return false; // Archive dir can contain only log files
} else if (suffix == Slice(kRocksDbTFileExt) ||
suffix == Slice(kLevelDbTFileExt)) {
*type = kTableFile;
} else if (suffix == Slice(kRocksDBBlobFileExt)) {
*type = kBlobFile;
} else if (suffix == Slice(kTempFileNameSuffix)) {
*type = kTempFile;
} else {
return false;
}
*number = num;
}
return true;
}
IOStatus SetCurrentFile(FileSystem* fs, const std::string& dbname,
uint64_t descriptor_number,
FSDirectory* directory_to_fsync) {
// Remove leading "dbname/" and add newline to manifest file name
std::string manifest = DescriptorFileName(dbname, descriptor_number);
Slice contents = manifest;
assert(contents.starts_with(dbname + "/"));
contents.remove_prefix(dbname.size() + 1);
std::string tmp = TempFileName(dbname, descriptor_number);
IOStatus s = WriteStringToFile(fs, contents.ToString() + "\n", tmp, true);
TEST_SYNC_POINT_CALLBACK("SetCurrentFile:BeforeRename", &s);
if (s.ok()) {
TEST_KILL_RANDOM("SetCurrentFile:0", rocksdb_kill_odds * REDUCE_ODDS2);
s = fs->RenameFile(tmp, CurrentFileName(dbname), IOOptions(), nullptr);
TEST_KILL_RANDOM("SetCurrentFile:1", rocksdb_kill_odds * REDUCE_ODDS2);
TEST_SYNC_POINT_CALLBACK("SetCurrentFile:AfterRename", &s);
}
if (s.ok()) {
if (directory_to_fsync != nullptr) {
s = directory_to_fsync->Fsync(IOOptions(), nullptr);
}
} else {
fs->DeleteFile(tmp, IOOptions(), nullptr);
}
return s;
}
Status SetIdentityFile(Env* env, const std::string& dbname,
const std::string& db_id) {
std::string id;
if (db_id.empty()) {
id = env->GenerateUniqueId();
} else {
id = db_id;
}
assert(!id.empty());
// Reserve the filename dbname/000000.dbtmp for the temporary identity file
std::string tmp = TempFileName(dbname, 0);
Status s = WriteStringToFile(env, id, tmp, true);
if (s.ok()) {
s = env->RenameFile(tmp, IdentityFileName(dbname));
}
if (!s.ok()) {
env->DeleteFile(tmp).PermitUncheckedError();
}
return s;
}
IOStatus SyncManifest(const ImmutableDBOptions* db_options,
WritableFileWriter* file) {
TEST_KILL_RANDOM("SyncManifest:0", rocksdb_kill_odds * REDUCE_ODDS2);
StopWatch sw(db_options->clock, db_options->statistics.get(),
MANIFEST_FILE_SYNC_MICROS);
return file->Sync(db_options->use_fsync);
}
Status GetInfoLogFiles(const std::shared_ptr<FileSystem>& fs,
const std::string& db_log_dir, const std::string& dbname,
std::string* parent_dir,
std::vector<std::string>* info_log_list) {
assert(parent_dir != nullptr);
assert(info_log_list != nullptr);
uint64_t number = 0;
FileType type = kWalFile;
if (!db_log_dir.empty()) {
*parent_dir = db_log_dir;
} else {
*parent_dir = dbname;
}
InfoLogPrefix info_log_prefix(!db_log_dir.empty(), dbname);
std::vector<std::string> file_names;
Status s = fs->GetChildren(*parent_dir, IOOptions(), &file_names, nullptr);
if (!s.ok()) {
return s;
}
for (auto& f : file_names) {
if (ParseFileName(f, &number, info_log_prefix.prefix, &type) &&
(type == kInfoLogFile)) {
info_log_list->push_back(f);
}
}
return Status::OK();
}
std::string NormalizePath(const std::string& path) {
std::string dst;
for (auto c : path) {
if (!dst.empty() && (c == kFilePathSeparator || c == '/') &&
(dst.back() == kFilePathSeparator || dst.back() == '/')) {
continue;
}
dst.push_back(c);
}
return dst;
}
} // namespace ROCKSDB_NAMESPACE