Skip to content

T
TDengine

taosdata / TDengine
1 年多 前同步成功

通知 1185
Star 22016
Fork 4786
T
TDengine
提交 0d2d3349 编写于 作者: X Xiaoyu Wang
浏览文件
操作

Merge remote-tracking branch 'origin/main' into fix/main_bugfix_wxy

上级 ec159239 0fad949c
展开全部 显示空白变更内容
内联 并排
Showing with 2029 addition and 2201 deletion
cmake/taosadapter_CMakeLists.txt.in
浏览文件 @ 0d2d3349
......@@ -2,7 +2,7 @@
# taosadapter
ExternalProject_Add(taosadapter
GIT_REPOSITORY https://github.com/taosdata/taosadapter.git
GIT_TAG 213f8b3
GIT_TAG 3e08996
SOURCE_DIR "${TD_SOURCE_DIR}/tools/taosadapter"
BINARY_DIR ""
#BUILD_IN_SOURCE TRUE
......
cmake/taostools_CMakeLists.txt.in
浏览文件 @ 0d2d3349
......@@ -2,7 +2,7 @@
# taos-tools
ExternalProject_Add(taos-tools
GIT_REPOSITORY https://github.com/taosdata/taos-tools.git
GIT_TAG 7d24ed5
GIT_TAG 0cd564a
SOURCE_DIR "${TD_SOURCE_DIR}/tools/taos-tools"
BINARY_DIR ""
#BUILD_IN_SOURCE TRUE
......
docs/examples/go/go.mod 已删除 100644 → 0
浏览文件 @ ec159239
module goexample
go 1.17
require github.com/taosdata/driver-go/v3 3.0
docs/examples/python/conn_native_pandas.py
浏览文件 @ 0d2d3349
import pandas
from sqlalchemy import create_engine
from sqlalchemy import create_engine, text
engine = create_engine("taos://root:taosdata@localhost:6030/power")
df = pandas.read_sql("SELECT * FROM meters", engine)
conn = engine.connect()
df = pandas.read_sql(text("SELECT * FROM power.meters"), conn)
conn.close()
# print index
print(df.index)
......
docs/examples/python/conn_rest_pandas.py
浏览文件 @ 0d2d3349
import pandas
from sqlalchemy import create_engine
from sqlalchemy import create_engine, text
engine = create_engine("taosrest://root:taosdata@localhost:6041")
df: pandas.DataFrame = pandas.read_sql("SELECT * FROM power.meters", engine)
conn = engine.connect()
df: pandas.DataFrame = pandas.read_sql(text("SELECT * FROM power.meters"), conn)
conn.close()
# print index
print(df.index)
......
docs/examples/python/connect_rest_examples.py
浏览文件 @ 0d2d3349
# ANCHOR: connect
from taosrest import connect, TaosRestConnection, TaosRestCursor
conn: TaosRestConnection = connect(url="http://localhost:6041",
conn = connect(url="http://localhost:6041",
user="root",
password="taosdata",
timeout=30)
......@@ -9,16 +9,17 @@ conn: TaosRestConnection = connect(url="http://localhost:6041",
# ANCHOR_END: connect
# ANCHOR: basic
# create STable
cursor: TaosRestCursor = conn.cursor()
cursor = conn.cursor()
cursor.execute("DROP DATABASE IF EXISTS power")
cursor.execute("CREATE DATABASE power")
cursor.execute("CREATE STABLE power.meters (ts TIMESTAMP, current FLOAT, voltage INT, phase FLOAT) TAGS (location BINARY(64), groupId INT)")
cursor.execute(
"CREATE STABLE power.meters (ts TIMESTAMP, current FLOAT, voltage INT, phase FLOAT) TAGS (location BINARY(64), groupId INT)")
# insert data
cursor.execute("""INSERT INTO power.d1001 USING power.meters TAGS(California.SanFrancisco, 2) VALUES ('2018-10-03 14:38:05.000', 10.30000, 219, 0.31000) ('2018-10-03 14:38:15.000', 12.60000, 218, 0.33000) ('2018-10-03 14:38:16.800', 12.30000, 221, 0.31000)
power.d1002 USING power.meters TAGS(California.SanFrancisco, 3) VALUES ('2018-10-03 14:38:16.650', 10.30000, 218, 0.25000)
power.d1003 USING power.meters TAGS(California.LosAngeles, 2) VALUES ('2018-10-03 14:38:05.500', 11.80000, 221, 0.28000) ('2018-10-03 14:38:16.600', 13.40000, 223, 0.29000)
power.d1004 USING power.meters TAGS(California.LosAngeles, 3) VALUES ('2018-10-03 14:38:05.000', 10.80000, 223, 0.29000) ('2018-10-03 14:38:06.500', 11.50000, 221, 0.35000)""")
cursor.execute("""INSERT INTO power.d1001 USING power.meters TAGS('California.SanFrancisco', 2) VALUES ('2018-10-03 14:38:05.000', 10.30000, 219, 0.31000) ('2018-10-03 14:38:15.000', 12.60000, 218, 0.33000) ('2018-10-03 14:38:16.800', 12.30000, 221, 0.31000)
power.d1002 USING power.meters TAGS('California.SanFrancisco', 3) VALUES ('2018-10-03 14:38:16.650', 10.30000, 218, 0.25000)
power.d1003 USING power.meters TAGS('California.LosAngeles', 2) VALUES ('2018-10-03 14:38:05.500', 11.80000, 221, 0.28000) ('2018-10-03 14:38:16.600', 13.40000, 223, 0.29000)
power.d1004 USING power.meters TAGS('California.LosAngeles', 3) VALUES ('2018-10-03 14:38:05.000', 10.80000, 223, 0.29000) ('2018-10-03 14:38:06.500', 11.50000, 221, 0.35000)""")
print("inserted row count:", cursor.rowcount)
# query data
......@@ -28,7 +29,7 @@ print("queried row count:", cursor.rowcount)
# get column names from cursor
column_names = [meta[0] for meta in cursor.description]
# get rows
data: list[tuple] = cursor.fetchall()
data = cursor.fetchall()
print(column_names)
for row in data:
print(row)
......
docs/examples/python/connection_usage_native_reference.py
浏览文件 @ 0d2d3349
......@@ -8,7 +8,7 @@ conn.execute("CREATE DATABASE test")
# change database. same as execute "USE db"
conn.select_db("test")
conn.execute("CREATE STABLE weather(ts TIMESTAMP, temperature FLOAT) TAGS (location INT)")
affected_row: int = conn.execute("INSERT INTO t1 USING weather TAGS(1) VALUES (now, 23.5) (now+1m, 23.5) (now+2m 24.4)")
affected_row = conn.execute("INSERT INTO t1 USING weather TAGS(1) VALUES (now, 23.5) (now+1m, 23.5) (now+2m, 24.4)")
print("affected_row", affected_row)
# output:
# affected_row 3
......@@ -16,10 +16,10 @@ print("affected_row", affected_row)
# ANCHOR: query
# Execute a sql and get its result set. It's useful for SELECT statement
result: taos.TaosResult = conn.query("SELECT * from weather")
result = conn.query("SELECT * from weather")
# Get fields from result
fields: taos.field.TaosFields = result.fields
fields = result.fields
for field in fields:
print(field) # {name: ts, type: 9, bytes: 8}
......
docs/examples/python/fast_write_example.py
浏览文件 @ 0d2d3349
# install dependencies:
# recommend python >= 3.8
# pip3 install faster-fifo
#
import logging
import math
import multiprocessing
import sys
import time
import os
from multiprocessing import Process
from faster_fifo import Queue
from multiprocessing import Process, Queue
from mockdatasource import MockDataSource
from queue import Empty
from typing import List
......@@ -22,8 +21,7 @@ TABLE_COUNT = 1000
QUEUE_SIZE = 1000000
MAX_BATCH_SIZE = 3000
read_processes = []
write_processes = []
_DONE_MESSAGE = '__DONE__'
def get_connection():
......@@ -44,41 +42,64 @@ def get_connection():
# ANCHOR: read
def run_read_task(task_id: int, task_queues: List[Queue]):
def run_read_task(task_id: int, task_queues: List[Queue], infinity):
table_count_per_task = TABLE_COUNT // READ_TASK_COUNT
data_source = MockDataSource(f"tb{task_id}", table_count_per_task)
data_source = MockDataSource(f"tb{task_id}", table_count_per_task, infinity)
try:
for batch in data_source:
if isinstance(batch, tuple):
batch = [batch]
for table_id, rows in batch:
# hash data to different queue
i = table_id % len(task_queues)
# block putting forever when the queue is full
task_queues[i].put_many(rows, block=True, timeout=-1)
for row in rows:
task_queues[i].put(row)
if not infinity:
for queue in task_queues:
queue.put(_DONE_MESSAGE)
except KeyboardInterrupt:
pass
finally:
logging.info('read task over')
# ANCHOR_END: read
# ANCHOR: write
def run_write_task(task_id: int, queue: Queue):
def run_write_task(task_id: int, queue: Queue, done_queue: Queue):
from sql_writer import SQLWriter
log = logging.getLogger(f"WriteTask-{task_id}")
writer = SQLWriter(get_connection)
lines = None
try:
while True:
over = False
lines = []
for _ in range(MAX_BATCH_SIZE):
try:
# get as many as possible
lines = queue.get_many(block=False, max_messages_to_get=MAX_BATCH_SIZE)
writer.process_lines(lines)
line = queue.get_nowait()
if line == _DONE_MESSAGE:
over = True
break
if line:
lines.append(line)
except Empty:
time.sleep(0.01)
time.sleep(0.1)
if len(lines) > 0:
writer.process_lines(lines)
if over:
done_queue.put(_DONE_MESSAGE)
break
except KeyboardInterrupt:
pass
except BaseException as e:
log.debug(f"lines={lines}")
raise e
finally:
writer.close()
log.debug('write task over')
# ANCHOR_END: write
......@@ -103,13 +124,11 @@ def set_global_config():
# ANCHOR: monitor
def run_monitor_process():
def run_monitor_process(done_queue: Queue):
log = logging.getLogger("DataBaseMonitor")
conn = None
try:
conn = get_connection()
conn.execute("DROP DATABASE IF EXISTS test")
conn.execute("CREATE DATABASE test")
conn.execute("CREATE STABLE test.meters (ts TIMESTAMP, current FLOAT, voltage INT, phase FLOAT) "
"TAGS (location BINARY(64), groupId INT)")
def get_count():
res = conn.query("SELECT count(*) FROM test.meters")
......@@ -118,32 +137,51 @@ def run_monitor_process():
last_count = 0
while True:
try:
done = done_queue.get_nowait()
if done == _DONE_MESSAGE:
break
except Empty:
pass
time.sleep(10)
count = get_count()
log.info(f"count={count} speed={(count - last_count) / 10}")
last_count = count
finally:
conn.close()
# ANCHOR_END: monitor
# ANCHOR: main
def main():
def main(infinity):
set_global_config()
logging.info(f"READ_TASK_COUNT={READ_TASK_COUNT}, WRITE_TASK_COUNT={WRITE_TASK_COUNT}, "
f"TABLE_COUNT={TABLE_COUNT}, QUEUE_SIZE={QUEUE_SIZE}, MAX_BATCH_SIZE={MAX_BATCH_SIZE}")
monitor_process = Process(target=run_monitor_process)
conn = get_connection()
conn.execute("DROP DATABASE IF EXISTS test")
conn.execute("CREATE DATABASE IF NOT EXISTS test")
conn.execute("CREATE STABLE IF NOT EXISTS test.meters (ts TIMESTAMP, current FLOAT, voltage INT, phase FLOAT) "
"TAGS (location BINARY(64), groupId INT)")
conn.close()
done_queue = Queue()
monitor_process = Process(target=run_monitor_process, args=(done_queue,))
monitor_process.start()
time.sleep(3) # waiting for database ready.
logging.debug(f"monitor task started with pid {monitor_process.pid}")
task_queues: List[Queue] = []
write_processes = []
read_processes = []
# create task queues
for i in range(WRITE_TASK_COUNT):
queue = Queue(max_size_bytes=QUEUE_SIZE)
queue = Queue()
task_queues.append(queue)
# create write processes
for i in range(WRITE_TASK_COUNT):
p = Process(target=run_write_task, args=(i, task_queues[i]))
p = Process(target=run_write_task, args=(i, task_queues[i], done_queue))
p.start()
logging.debug(f"WriteTask-{i} started with pid {p.pid}")
write_processes.append(p)
......@@ -151,13 +189,19 @@ def main():
# create read processes
for i in range(READ_TASK_COUNT):
queues = assign_queues(i, task_queues)
p = Process(target=run_read_task, args=(i, queues))
p = Process(target=run_read_task, args=(i, queues, infinity))
p.start()
logging.debug(f"ReadTask-{i} started with pid {p.pid}")
read_processes.append(p)
try:
monitor_process.join()
for p in read_processes:
p.join()
for p in write_processes:
p.join()
time.sleep(1)
return
except KeyboardInterrupt:
monitor_process.terminate()
[p.terminate() for p in read_processes]
......@@ -176,5 +220,6 @@ def assign_queues(read_task_id, task_queues):
if __name__ == '__main__':
main()
multiprocessing.set_start_method('spawn')
main(False)
# ANCHOR_END: main
docs/examples/python/kafka_example.py
浏览文件 @ 0d2d3349
......@@ -26,7 +26,8 @@ class Consumer(object):
'bath_consume': True,
'batch_size': 1000,
'async_model': True,
'workers': 10
'workers': 10,
'testing': False
}
LOCATIONS = ['California.SanFrancisco', 'California.LosAngles', 'California.SanDiego', 'California.SanJose',
......@@ -46,6 +47,7 @@ class Consumer(object):
def __init__(self, **configs):
self.config: dict = self.DEFAULT_CONFIGS
self.config.update(configs)
if not self.config.get('testing'):
self.consumer = KafkaConsumer(
self.config.get('kafka_topic'), # topic
bootstrap_servers=self.config.get('kafka_brokers'),
......@@ -60,7 +62,7 @@ class Consumer(object):
)
if self.config.get('async_model'):
self.pool = ThreadPoolExecutor(max_workers=self.config.get('workers'))
self.tasks: list[Future] = []
self.tasks = []
# tags and table mapping # key: {location}_{groupId} value:
self.tag_table_mapping = {}
i = 0
......@@ -115,14 +117,14 @@ class Consumer(object):
if self.taos is not None:
self.taos.close()
def _run(self, f: Callable[[ConsumerRecord], bool]):
def _run(self, f):
for message in self.consumer:
if self.config.get('async_model'):
self.pool.submit(f(message))
else:
f(message)
def _run_batch(self, f: Callable[[list[list[ConsumerRecord]]], None]):
def _run_batch(self, f):
while True:
messages = self.consumer.poll(timeout_ms=500, max_records=self.config.get('batch_size'))
if messages:
......@@ -140,7 +142,7 @@ class Consumer(object):
logging.info('## insert sql %s', sql)
return self.taos.execute(sql=sql) == 1
def _to_taos_batch(self, messages: list[list[ConsumerRecord]]):
def _to_taos_batch(self, messages):
sql = self._build_sql_batch(messages=messages)
if len(sql) == 0: # decode error, skip
return
......@@ -162,7 +164,7 @@ class Consumer(object):
table_name = self._get_table_name(location=location, group_id=group_id)
return self.INSERT_PART_SQL.format(table_name, ts, current, voltage, phase)
def _build_sql_batch(self, messages: list[list[ConsumerRecord]]) -> str:
def _build_sql_batch(self, messages) -> str:
sql_list = []
for partition_messages in messages:
for message in partition_messages:
......@@ -186,7 +188,55 @@ def _get_location_and_group(key: str) -> (str, int):
return fields[0], fields[1]
def test_to_taos(consumer: Consumer):
msg = {
'location': 'California.SanFrancisco',
'groupId': 1,
'ts': '2022-12-06 15:13:38.643',
'current': 3.41,
'voltage': 105,
'phase': 0.02027,
}
record = ConsumerRecord(checksum=None, headers=None, offset=1, key=None, value=json.dumps(msg), partition=1,
topic='test', serialized_key_size=None, serialized_header_size=None,
serialized_value_size=None, timestamp=time.time(), timestamp_type=None)
assert consumer._to_taos(message=record)
def test_to_taos_batch(consumer: Consumer):
records = [
[
ConsumerRecord(checksum=None, headers=None, offset=1, key=None,
value=json.dumps({'location': 'California.SanFrancisco',
'groupId': 1,
'ts': '2022-12-06 15:13:38.643',
'current': 3.41,
'voltage': 105,
'phase': 0.02027, }),
partition=1, topic='test', serialized_key_size=None, serialized_header_size=None,
serialized_value_size=None, timestamp=time.time(), timestamp_type=None),
ConsumerRecord(checksum=None, headers=None, offset=1, key=None,
value=json.dumps({'location': 'California.LosAngles',
'groupId': 2,
'ts': '2022-12-06 15:13:39.643',
'current': 3.41,
'voltage': 102,
'phase': 0.02027, }),
partition=1, topic='test', serialized_key_size=None, serialized_header_size=None,
serialized_value_size=None, timestamp=time.time(), timestamp_type=None),
]
]
consumer._to_taos_batch(messages=records)
if __name__ == '__main__':
consumer = Consumer(async_model=True)
consumer = Consumer(async_model=True, testing=True)
# init env
consumer.init_env()
consumer.consume()
\ No newline at end of file
# consumer.consume()
# test build sql
# test build sql batch
test_to_taos(consumer)
test_to_taos_batch(consumer)
\ No newline at end of file
docs/examples/python/mockdatasource.py
浏览文件 @ 0d2d3349
......@@ -10,13 +10,14 @@ class MockDataSource:
"9.4,118,0.141,California.SanFrancisco,4"
]
def __init__(self, tb_name_prefix, table_count):
def __init__(self, tb_name_prefix, table_count, infinity=True):
self.table_name_prefix = tb_name_prefix + "_"
self.table_count = table_count
self.max_rows = 10000000
self.current_ts = round(time.time() * 1000) - self.max_rows * 100
# [(tableId, tableName, values),]
self.data = self._init_data()
self.infinity = infinity
def _init_data(self):
lines = self.samples * (self.table_count // 5 + 1)
......@@ -28,6 +29,9 @@ class MockDataSource:
def __iter__(self):
self.row = 0
if not self.infinity:
return iter(self._iter_data())
else:
return self
def __next__(self):
......@@ -35,7 +39,9 @@ class MockDataSource:
next 1000 rows for each table.
return: {tableId:[row,...]}
"""
# generate 1000 timestamps
return self._iter_data()
def _iter_data(self):
ts = []
for _ in range(1000):
self.current_ts += 100
......@@ -47,3 +53,10 @@ class MockDataSource:
rows = [table_name + ',' + t + ',' + values for t in ts]
result.append((table_id, rows))
return result
if __name__ == '__main__':
datasource = MockDataSource('t', 10, False)
for data in datasource:
print(data)
\ No newline at end of file
docs/examples/python/sql_writer.py
浏览文件 @ 0d2d3349
......@@ -10,6 +10,7 @@ class SQLWriter:
self._tb_tags = {}
self._conn = get_connection_func()
self._max_sql_length = self.get_max_sql_length()
self._conn.execute("create database if not exists test")
self._conn.execute("USE test")
def get_max_sql_length(self):
......@@ -20,7 +21,7 @@ class SQLWriter:
return int(r[1])
return 1024 * 1024
def process_lines(self, lines: str):
def process_lines(self, lines: [str]):
"""
:param lines: [[tbName,ts,current,voltage,phase,location,groupId]]
"""
......@@ -60,6 +61,7 @@ class SQLWriter:
buf.append(q)
sql_len += len(q)
sql += " ".join(buf)
self.create_tables()
self.execute_sql(sql)
self._tb_values.clear()
......@@ -88,3 +90,23 @@ class SQLWriter:
except BaseException as e:
self.log.error("Execute SQL: %s", sql)
raise e
def close(self):
if self._conn:
self._conn.close()
if __name__ == '__main__':
def get_connection_func():
conn = taos.connect()
return conn
writer = SQLWriter(get_connection_func=get_connection_func)
writer.execute_sql(
"create stable if not exists meters (ts timestamp, current float, voltage int, phase float) "
"tags (location binary(64), groupId int)")
writer.execute_sql(
"INSERT INTO d21001 USING meters TAGS ('California.SanFrancisco', 2) "
"VALUES ('2021-07-13 14:06:32.272', 10.2, 219, 0.32)")
\ No newline at end of file
docs/examples/python/tmq_example.py
浏览文件 @ 0d2d3349
from taos.tmq import Consumer
import taos
from taos.tmq import *
conn = taos.connect()
print("init")
conn.execute("drop topic if exists topic_ctb_column")
conn.execute("drop database if exists py_tmq")
conn.execute("create database if not exists py_tmq vgroups 2")
conn.select_db("py_tmq")
conn.execute(
"create stable if not exists stb1 (ts timestamp, c1 int, c2 float, c3 binary(10)) tags(t1 int)"
)
conn.execute("create table if not exists tb1 using stb1 tags(1)")
conn.execute("create table if not exists tb2 using stb1 tags(2)")
conn.execute("create table if not exists tb3 using stb1 tags(3)")
print("create topic")
conn.execute(
"create topic if not exists topic_ctb_column as select ts, c1, c2, c3 from stb1"
)
print("build consumer")
conf = TaosTmqConf()
conf.set("group.id", "tg2")
conf.set("td.connect.user", "root")
conf.set("td.connect.pass", "taosdata")
conf.set("enable.auto.commit", "true")
def tmq_commit_cb_print(tmq, resp, offset, param=None):
print(f"commit: {resp}, tmq: {tmq}, offset: {offset}, param: {param}")
conf.set_auto_commit_cb(tmq_commit_cb_print, None)
tmq = conf.new_consumer()
print("build topic list")
topic_list = TaosTmqList()
topic_list.append("topic_ctb_column")
print("basic consume loop")
tmq.subscribe(topic_list)
sub_list = tmq.subscription()
print("subscribed topics: ", sub_list)
while 1:
res = tmq.poll(1000)
if res:
topic = res.get_topic_name()
vg = res.get_vgroup_id()
db = res.get_db_name()
print(f"topic: {topic}\nvgroup id: {vg}\ndb: {db}")
for row in res:
print(row)
def init_tmq_env(db, topic):
conn = taos.connect()
conn.execute("drop topic if exists {}".format(topic))
conn.execute("drop database if exists {}".format(db))
conn.execute("create database if not exists {}".format(db))
conn.select_db(db)
conn.execute(
"create stable if not exists stb1 (ts timestamp, c1 int, c2 float, c3 varchar(16)) tags(t1 int, t3 varchar(16))")
conn.execute("create table if not exists tb1 using stb1 tags(1, 't1')")
conn.execute("create table if not exists tb2 using stb1 tags(2, 't2')")
conn.execute("create table if not exists tb3 using stb1 tags(3, 't3')")
conn.execute("create topic if not exists {} as select ts, c1, c2, c3 from stb1".format(topic))
conn.execute("insert into tb1 values (now, 1, 1.0, 'tmq test')")
conn.execute("insert into tb2 values (now, 2, 2.0, 'tmq test')")
conn.execute("insert into tb3 values (now, 3, 3.0, 'tmq test')")
def cleanup(db, topic):
conn = taos.connect()
conn.execute("drop topic if exists {}".format(topic))
conn.execute("drop database if exists {}".format(db))
if __name__ == '__main__':
init_tmq_env("tmq_test", "tmq_test_topic") # init env
consumer = Consumer(
{
"group.id": "tg2",
"td.connect.user": "root",
"td.connect.pass": "taosdata",
"enable.auto.commit": "true",
}
)
consumer.subscribe(["tmq_test_topic"])
try:
while True:
res = consumer.poll(1)
if not res:
break
err = res.error()
if err is not None:
raise err
val = res.value()
for block in val:
print(block.fetchall())
finally:
consumer.unsubscribe()
consumer.close()
cleanup("tmq_test", "tmq_test_topic")
\ No newline at end of file
include/libs/wal/wal.h
浏览文件 @ 0d2d3349
......@@ -201,6 +201,7 @@ int32_t walFetchHead(SWalReader *pRead, int64_t ver, SWalCkHead *pHead);
int32_t walFetchBody(SWalReader *pRead, SWalCkHead **ppHead);
int32_t walSkipFetchBody(SWalReader *pRead, const SWalCkHead *pHead);
SWalRef *walRefFirstVer(SWal *, SWalRef *);
SWalRef *walRefCommittedVer(SWal *);
SWalRef *walOpenRef(SWal *);
......
packaging/tools/install.sh
浏览文件 @ 0d2d3349
......@@ -210,8 +210,8 @@ function install_bin() {
[ -x ${install_main_dir}/bin/${serverName} ] && ${csudo}ln -s ${install_main_dir}/bin/${serverName} ${bin_link_dir}/${serverName} || :
[ -x ${install_main_dir}/bin/${udfdName} ] && ${csudo}ln -s ${install_main_dir}/bin/${udfdName} ${bin_link_dir}/${udfdName} || :
[ -x ${install_main_dir}/bin/${adapterName} ] && ${csudo}ln -s ${install_main_dir}/bin/${adapterName} ${bin_link_dir}/${adapterName} || :
[ -x ${install_main_dir}/bin/${benchmarkName} ] && ${csudo}ln -s ${install_main_dir}/bin/${benchmarkName} ${bin_link_dir}/${demoName} || :
[ -x ${install_main_dir}/bin/${benchmarkName} ] && ${csudo}ln -s ${install_main_dir}/bin/${benchmarkName} ${bin_link_dir}/${benchmarkName} || :
[ -x ${install_main_dir}/bin/${benchmarkName} ] && ${csudo}ln -sf ${install_main_dir}/bin/${benchmarkName} ${bin_link_dir}/${demoName} || :
[ -x ${install_main_dir}/bin/${benchmarkName} ] && ${csudo}ln -sf ${install_main_dir}/bin/${benchmarkName} ${bin_link_dir}/${benchmarkName} || :
[ -x ${install_main_dir}/bin/${dumpName} ] && ${csudo}ln -s ${install_main_dir}/bin/${dumpName} ${bin_link_dir}/${dumpName} || :
[ -x ${install_main_dir}/bin/${xname} ] && ${csudo}ln -s ${install_main_dir}/bin/${xname} ${bin_link_dir}/${xname} || :
[ -x ${install_main_dir}/bin/TDinsight.sh ] && ${csudo}ln -s ${install_main_dir}/bin/TDinsight.sh ${bin_link_dir}/TDinsight.sh || :
......@@ -746,7 +746,7 @@ function is_version_compatible() {
deb_erase() {
confirm=""
while [ "" == "${confirm}" ]; do
echo -e -n "${RED}Exist tdengine deb detected, do you want to remove it? [yes|no] ${NC}:"
echo -e -n "${RED}Existing TDengine deb is detected, do you want to remove it? [yes|no] ${NC}:"
read confirm
if [ "yes" == "$confirm" ]; then
${csudo}dpkg --remove tdengine ||:
......@@ -760,7 +760,7 @@ deb_erase() {
rpm_erase() {
confirm=""
while [ "" == "${confirm}" ]; do
echo -e -n "${RED}Exist tdengine rpm detected, do you want to remove it? [yes|no] ${NC}:"
echo -e -n "${RED}Existing TDengine rpm is detected, do you want to remove it? [yes|no] ${NC}:"
read confirm
if [ "yes" == "$confirm" ]; then
${csudo}rpm -e tdengine ||:
......@@ -787,7 +787,7 @@ function updateProduct() {
if echo $osinfo | grep -qwi "centos"; then
rpm -q tdengine 2>&1 > /dev/null && rpm_erase tdengine ||:
elif echo $osinfo | grep -qwi "ubuntu"; then
dpkg -l tdengine 2>&1 > /dev/null && deb_erase tdengine ||:
dpkg -l tdengine 2>&1 | grep ii > /dev/null && deb_erase tdengine ||:
fi
tar -zxf ${tarName}
......
source/dnode/mgmt/mgmt_vnode/src/vmInt.c
浏览文件 @ 0d2d3349
......@@ -79,8 +79,6 @@ int32_t vmOpenVnode(SVnodeMgmt *pMgmt, SWrapperCfg *pCfg, SVnode *pImpl) {
void vmCloseVnode(SVnodeMgmt *pMgmt, SVnodeObj *pVnode) {
char path[TSDB_FILENAME_LEN] = {0};
vnodeProposeCommitOnNeed(pVnode->pImpl);
taosThreadRwlockWrlock(&pMgmt->lock);
taosHashRemove(pMgmt->hash, &pVnode->vgId, sizeof(int32_t));
taosThreadRwlockUnlock(&pMgmt->lock);
......
source/dnode/vnode/src/inc/tsdb.h
浏览文件 @ 0d2d3349
......@@ -202,6 +202,7 @@ int32_t tsdbCmprColData(SColData *pColData, int8_t cmprAlg, SBlockCol *pBlockCol
uint8_t **ppBuf);
int32_t tsdbDecmprColData(uint8_t *pIn, SBlockCol *pBlockCol, int8_t cmprAlg, int32_t nVal, SColData *pColData,
uint8_t **ppBuf);
int32_t tRowInfoCmprFn(const void *p1, const void *p2);
// tsdbMemTable ==============================================================================================
// SMemTable
int32_t tsdbMemTableCreate(STsdb *pTsdb, SMemTable **ppMemTable);
......
source/dnode/vnode/src/inc/vnodeInt.h
浏览文件 @ 0d2d3349
......@@ -247,7 +247,7 @@ int32_t tsdbSnapReaderClose(STsdbSnapReader** ppReader);
int32_t tsdbSnapRead(STsdbSnapReader* pReader, uint8_t** ppData);
// STsdbSnapWriter ========================================
int32_t tsdbSnapWriterOpen(STsdb* pTsdb, int64_t sver, int64_t ever, STsdbSnapWriter** ppWriter);
int32_t tsdbSnapWrite(STsdbSnapWriter* pWriter, uint8_t* pData, uint32_t nData);
int32_t tsdbSnapWrite(STsdbSnapWriter* pWriter, SSnapDataHdr* pHdr);
int32_t tsdbSnapWriterPrepareClose(STsdbSnapWriter* pWriter);
int32_t tsdbSnapWriterClose(STsdbSnapWriter** ppWriter, int8_t rollback);
// STqSnapshotReader ==
......
source/dnode/vnode/src/sma/smaSnapshot.c
浏览文件 @ 0d2d3349
......@@ -423,10 +423,10 @@ int32_t rsmaSnapWrite(SRSmaSnapWriter* pWriter, uint8_t* pData, uint32_t nData)
// rsma1/rsma2
if (pHdr->type == SNAP_DATA_RSMA1) {
pHdr->type = SNAP_DATA_TSDB;
code = tsdbSnapWrite(pWriter->pDataWriter[0], pData, nData);
code = tsdbSnapWrite(pWriter->pDataWriter[0], pHdr);
} else if (pHdr->type == SNAP_DATA_RSMA2) {
pHdr->type = SNAP_DATA_TSDB;
code = tsdbSnapWrite(pWriter->pDataWriter[1], pData, nData);
code = tsdbSnapWrite(pWriter->pDataWriter[1], pHdr);
} else if (pHdr->type == SNAP_DATA_QTASK) {
code = rsmaSnapWriteQTaskInfo(pWriter, pData, nData);
} else {
......
source/dnode/vnode/src/tq/tq.c
浏览文件 @ 0d2d3349
......@@ -521,7 +521,12 @@ int32_t tqProcessPollReq(STQ* pTq, SRpcMsg* pMsg) {
tqOffsetResetToData(&fetchOffsetNew, 0, 0);
}
} else {
tqOffsetResetToLog(&fetchOffsetNew, walGetFirstVer(pTq->pVnode->pWal));
pHandle->pRef = walRefFirstVer(pTq->pVnode->pWal, pHandle->pRef);
if (pHandle->pRef == NULL) {
terrno = TSDB_CODE_OUT_OF_MEMORY;
return -1;
}
tqOffsetResetToLog(&fetchOffsetNew, pHandle->pRef->refVer - 1);
}
} else if (reqOffset.type == TMQ_OFFSET__RESET_LATEST) {
if (pHandle->execHandle.subType == TOPIC_SUB_TYPE__COLUMN) {
......@@ -719,6 +724,8 @@ int32_t tqProcessPollReq(STQ* pTq, SRpcMsg* pMsg) {
int32_t tqProcessDeleteSubReq(STQ* pTq, int64_t version, char* msg, int32_t msgLen) {
SMqVDeleteReq* pReq = (SMqVDeleteReq*)msg;
tqDebug("vgId:%d, delete sub: %s", pTq->pVnode->config.vgId, pReq->subKey);
taosWLockLatch(&pTq->pushLock);
int32_t code = taosHashRemove(pTq->pPushMgr, pReq->subKey, strlen(pReq->subKey));
if (code != 0) {
......
source/dnode/vnode/src/vnd/vnodeSnapshot.c
浏览文件 @ 0d2d3349
......@@ -455,7 +455,7 @@ int32_t vnodeSnapWrite(SVSnapWriter *pWriter, uint8_t *pData, uint32_t nData) {
if (code) goto _err;
}
code = tsdbSnapWrite(pWriter->pTsdbSnapWriter, pData, nData);
code = tsdbSnapWrite(pWriter->pTsdbSnapWriter, pHdr);
if (code) goto _err;
} break;
case SNAP_DATA_TQ_HANDLE: {
......
source/libs/executor/src/timewindowoperator.c
浏览文件 @ 0d2d3349
......@@ -2477,7 +2477,19 @@ static void doStreamIntervalAggImpl(SOperatorInfo* pOperatorInfo, SSDataBlock* p
pInfo->delKey = key;
}
int32_t prevEndPos = (forwardRows - 1) * step + startPos;
ASSERT(pSDataBlock->info.window.skey > 0 && pSDataBlock->info.window.ekey > 0);
if (pSDataBlock->info.window.skey <= 0 || pSDataBlock->info.window.ekey <= 0) {
qError("table uid %" PRIu64 " data block timestamp range may not be calculated! minKey %" PRId64
",maxKey %" PRId64,
pSDataBlock->info.id.uid, pSDataBlock->info.window.skey, pSDataBlock->info.window.ekey);
blockDataUpdateTsWindow(pSDataBlock, 0);
// timestamp of the data is incorrect
if (pSDataBlock->info.window.skey <= 0 || pSDataBlock->info.window.ekey <= 0) {
qError("table uid %" PRIu64 " data block timestamp is out of range! minKey %" PRId64 ",maxKey %" PRId64,
pSDataBlock->info.id.uid, pSDataBlock->info.window.skey, pSDataBlock->info.window.ekey);
}
}
if (IS_FINAL_OP(pInfo)) {
startPos = getNextQualifiedFinalWindow(&pInfo->interval, &nextWin, &pSDataBlock->info, tsCols, prevEndPos);
} else {
......
source/libs/qworker/inc/qwInt.h
浏览文件 @ 0d2d3349
......@@ -194,6 +194,8 @@ typedef struct SQWorker {
SMsgCb msgCb;
SQWStat stat;
int32_t *destroyed;
int8_t nodeStopped;
} SQWorker;
typedef struct SQWorkerMgmt {
......
source/libs/qworker/src/qwUtil.c
浏览文件 @ 0d2d3349
......@@ -213,9 +213,15 @@ int32_t qwAcquireTaskCtx(QW_FPARAMS_DEF, SQWTaskCtx **ctx) {
QW_SET_QTID(id, qId, tId, eId);
*ctx = taosHashAcquire(mgmt->ctxHash, id, sizeof(id));
int8_t nodeStopped = atomic_load_8(&mgmt->nodeStopped);
if (NULL == (*ctx)) {
if (!nodeStopped) {
QW_TASK_DLOG_E("task ctx not exist, may be dropped");
QW_ERR_RET(TSDB_CODE_QRY_TASK_CTX_NOT_EXIST);
} else {
QW_TASK_DLOG_E("node stopped");
QW_ERR_RET(TSDB_CODE_VND_STOPPED);
}
}
return TSDB_CODE_SUCCESS;
......@@ -226,9 +232,16 @@ int32_t qwGetTaskCtx(QW_FPARAMS_DEF, SQWTaskCtx **ctx) {
QW_SET_QTID(id, qId, tId, eId);
*ctx = taosHashGet(mgmt->ctxHash, id, sizeof(id));
int8_t nodeStopped = atomic_load_8(&mgmt->nodeStopped);
if (NULL == (*ctx)) {
if (!nodeStopped) {
QW_TASK_DLOG_E("task ctx not exist, may be dropped");
QW_ERR_RET(TSDB_CODE_QRY_TASK_CTX_NOT_EXIST);
} else {
QW_TASK_DLOG_E("node stopped");
QW_ERR_RET(TSDB_CODE_VND_STOPPED);
}
}
return TSDB_CODE_SUCCESS;
......
source/libs/qworker/src/qworker.c
浏览文件 @ 0d2d3349
......@@ -1188,6 +1188,9 @@ void qWorkerStopAllTasks(void *qWorkerMgmt) {
uint64_t qId, tId, sId;
int32_t eId;
int64_t rId = 0;
atomic_store_8(&mgmt->nodeStopped, 1);
void *pIter = taosHashIterate(mgmt->ctxHash, NULL);
while (pIter) {
SQWTaskCtx *ctx = (SQWTaskCtx *)pIter;
......
source/libs/stream/src/streamMeta.c
浏览文件 @ 0d2d3349
......@@ -207,6 +207,7 @@ void streamMetaRemoveTask(SStreamMeta* pMeta, int32_t taskId) {
if (ppTask) {
SStreamTask* pTask = *ppTask;
taosHashRemove(pMeta->pTasks, &taskId, sizeof(int32_t));
tdbTbDelete(pMeta->pTaskDb, &taskId, sizeof(int32_t), pMeta->txn);
/*if (pTask->timer) {
* taosTmrStop(pTask->timer);*/
/*pTask->timer = NULL;*/
......
source/libs/sync/src/syncAppendEntries.c
浏览文件 @ 0d2d3349
......@@ -89,45 +89,6 @@
// /\ UNCHANGED <<candidateVars, leaderVars>>
//
int32_t syncNodeFollowerCommit(SSyncNode* ths, SyncIndex newCommitIndex) {
ASSERT(false && "deprecated");
if (ths->state != TAOS_SYNC_STATE_FOLLOWER) {
sNTrace(ths, "can not do follower commit");
return -1;
}
// maybe update commit index, leader notice me
if (newCommitIndex > ths->commitIndex) {
// has commit entry in local
if (newCommitIndex <= ths->pLogStore->syncLogLastIndex(ths->pLogStore)) {
// advance commit index to sanpshot first
SSnapshot snapshot;
ths->pFsm->FpGetSnapshotInfo(ths->pFsm, &snapshot);
if (snapshot.lastApplyIndex >= 0 && snapshot.lastApplyIndex > ths->commitIndex) {
SyncIndex commitBegin = ths->commitIndex;
SyncIndex commitEnd = snapshot.lastApplyIndex;
ths->commitIndex = snapshot.lastApplyIndex;
sNTrace(ths, "commit by snapshot from index:%" PRId64 " to index:%" PRId64, commitBegin, commitEnd);
}
SyncIndex beginIndex = ths->commitIndex + 1;
SyncIndex endIndex = newCommitIndex;
// update commit index
ths->commitIndex = newCommitIndex;
// call back Wal
int32_t code = ths->pLogStore->syncLogUpdateCommitIndex(ths->pLogStore, ths->commitIndex);
ASSERT(code == 0);
code = syncNodeDoCommit(ths, beginIndex, endIndex, ths->state);
ASSERT(code == 0);
}
}
return 0;
}
SSyncRaftEntry* syncBuildRaftEntryFromAppendEntries(const SyncAppendEntries* pMsg) {
SSyncRaftEntry* pEntry = taosMemoryMalloc(pMsg->dataLen);
if (pEntry == NULL) {
......@@ -232,256 +193,3 @@ _IGNORE:
rpcFreeCont(rpcRsp.pCont);
return 0;
}
int32_t syncNodeOnAppendEntriesOld(SSyncNode* ths, const SRpcMsg* pRpcMsg) {
SyncAppendEntries* pMsg = pRpcMsg->pCont;
SRpcMsg rpcRsp = {0};
// if already drop replica, do not process
if (!syncNodeInRaftGroup(ths, &(pMsg->srcId))) {
syncLogRecvAppendEntries(ths, pMsg, "not in my config");
goto _IGNORE;
}
// prepare response msg
int32_t code = syncBuildAppendEntriesReply(&rpcRsp, ths->vgId);
if (code != 0) {
syncLogRecvAppendEntries(ths, pMsg, "build rsp error");
goto _IGNORE;
}
SyncAppendEntriesReply* pReply = rpcRsp.pCont;
pReply->srcId = ths->myRaftId;
pReply->destId = pMsg->srcId;
pReply->term = ths->raftStore.currentTerm;
pReply->success = false;
// pReply->matchIndex = ths->pLogStore->syncLogLastIndex(ths->pLogStore);
pReply->matchIndex = SYNC_INDEX_INVALID;
pReply->lastSendIndex = pMsg->prevLogIndex + 1;
pReply->startTime = ths->startTime;
if (pMsg->term < ths->raftStore.currentTerm) {
syncLogRecvAppendEntries(ths, pMsg, "reject, small term");
goto _SEND_RESPONSE;
}
if (pMsg->term > ths->raftStore.currentTerm) {
pReply->term = pMsg->term;
}
syncNodeStepDown(ths, pMsg->term);
syncNodeResetElectTimer(ths);
SyncIndex startIndex = ths->pLogStore->syncLogBeginIndex(ths->pLogStore);
SyncIndex lastIndex = ths->pLogStore->syncLogLastIndex(ths->pLogStore);
if (pMsg->prevLogIndex > lastIndex) {
syncLogRecvAppendEntries(ths, pMsg, "reject, index not match");
goto _SEND_RESPONSE;
}
if (pMsg->prevLogIndex >= startIndex) {
SyncTerm myPreLogTerm = syncNodeGetPreTerm(ths, pMsg->prevLogIndex + 1);
// ASSERT(myPreLogTerm != SYNC_TERM_INVALID);
if (myPreLogTerm == SYNC_TERM_INVALID) {
syncLogRecvAppendEntries(ths, pMsg, "reject, pre-term invalid");
goto _SEND_RESPONSE;
}
if (myPreLogTerm != pMsg->prevLogTerm) {
syncLogRecvAppendEntries(ths, pMsg, "reject, pre-term not match");
goto _SEND_RESPONSE;
}
}
// accept
pReply->success = true;
bool hasAppendEntries = pMsg->dataLen > 0;
if (hasAppendEntries) {
SSyncRaftEntry* pAppendEntry = syncEntryBuildFromAppendEntries(pMsg);
ASSERT(pAppendEntry != NULL);
SyncIndex appendIndex = pMsg->prevLogIndex + 1;
LRUHandle* hLocal = NULL;
LRUHandle* hAppend = NULL;
int32_t code = 0;
SSyncRaftEntry* pLocalEntry = NULL;
SLRUCache* pCache = ths->pLogStore->pCache;
hLocal = taosLRUCacheLookup(pCache, &appendIndex, sizeof(appendIndex));
if (hLocal) {
pLocalEntry = (SSyncRaftEntry*)taosLRUCacheValue(pCache, hLocal);
code = 0;
ths->pLogStore->cacheHit++;
sNTrace(ths, "hit cache index:%" PRId64 ", bytes:%u, %p", appendIndex, pLocalEntry->bytes, pLocalEntry);
} else {
ths->pLogStore->cacheMiss++;
sNTrace(ths, "miss cache index:%" PRId64, appendIndex);
code = ths->pLogStore->syncLogGetEntry(ths->pLogStore, appendIndex, &pLocalEntry);
}
if (code == 0) {
// get local entry success
if (pLocalEntry->term == pAppendEntry->term) {
// do nothing
sNTrace(ths, "log match, do nothing, index:%" PRId64, appendIndex);
} else {
// truncate
code = ths->pLogStore->syncLogTruncate(ths->pLogStore, appendIndex);
if (code != 0) {
char logBuf[128];
snprintf(logBuf, sizeof(logBuf), "ignore, truncate error, append-index:%" PRId64, appendIndex);
syncLogRecvAppendEntries(ths, pMsg, logBuf);
if (hLocal) {
taosLRUCacheRelease(ths->pLogStore->pCache, hLocal, false);
} else {
syncEntryDestroy(pLocalEntry);
}
if (hAppend) {
taosLRUCacheRelease(ths->pLogStore->pCache, hAppend, false);
} else {
syncEntryDestroy(pAppendEntry);
}
goto _IGNORE;
}
ASSERT(pAppendEntry->index == appendIndex);
// append
code = ths->pLogStore->syncLogAppendEntry(ths->pLogStore, pAppendEntry, false);
if (code != 0) {
char logBuf[128];
snprintf(logBuf, sizeof(logBuf), "ignore, append error, append-index:%" PRId64, appendIndex);
syncLogRecvAppendEntries(ths, pMsg, logBuf);
if (hLocal) {
taosLRUCacheRelease(ths->pLogStore->pCache, hLocal, false);
} else {
syncEntryDestroy(pLocalEntry);
}
if (hAppend) {
taosLRUCacheRelease(ths->pLogStore->pCache, hAppend, false);
} else {
syncEntryDestroy(pAppendEntry);
}
goto _IGNORE;
}
syncCacheEntry(ths->pLogStore, pAppendEntry, &hAppend);
}
} else {
if (terrno == TSDB_CODE_WAL_LOG_NOT_EXIST) {
// log not exist
// truncate
code = ths->pLogStore->syncLogTruncate(ths->pLogStore, appendIndex);
if (code != 0) {
char logBuf[128];
snprintf(logBuf, sizeof(logBuf), "ignore, log not exist, truncate error, append-index:%" PRId64, appendIndex);
syncLogRecvAppendEntries(ths, pMsg, logBuf);
syncEntryDestroy(pLocalEntry);
syncEntryDestroy(pAppendEntry);
goto _IGNORE;
}
// append
code = ths->pLogStore->syncLogAppendEntry(ths->pLogStore, pAppendEntry, false);
if (code != 0) {
char logBuf[128];
snprintf(logBuf, sizeof(logBuf), "ignore, log not exist, append error, append-index:%" PRId64, appendIndex);
syncLogRecvAppendEntries(ths, pMsg, logBuf);
if (hLocal) {
taosLRUCacheRelease(ths->pLogStore->pCache, hLocal, false);
} else {
syncEntryDestroy(pLocalEntry);
}
if (hAppend) {
taosLRUCacheRelease(ths->pLogStore->pCache, hAppend, false);
} else {
syncEntryDestroy(pAppendEntry);
}
goto _IGNORE;
}
syncCacheEntry(ths->pLogStore, pAppendEntry, &hAppend);
} else {
// get local entry success
char logBuf[128];
snprintf(logBuf, sizeof(logBuf), "ignore, get local entry error, append-index:%" PRId64 " err:%d", appendIndex,
terrno);
syncLogRecvAppendEntries(ths, pMsg, logBuf);
if (hLocal) {
taosLRUCacheRelease(ths->pLogStore->pCache, hLocal, false);
} else {
syncEntryDestroy(pLocalEntry);
}
if (hAppend) {
taosLRUCacheRelease(ths->pLogStore->pCache, hAppend, false);
} else {
syncEntryDestroy(pAppendEntry);
}
goto _IGNORE;
}
}
// update match index
pReply->matchIndex = pAppendEntry->index;
if (hLocal) {
taosLRUCacheRelease(ths->pLogStore->pCache, hLocal, false);
} else {
syncEntryDestroy(pLocalEntry);
}
if (hAppend) {
taosLRUCacheRelease(ths->pLogStore->pCache, hAppend, false);
} else {
syncEntryDestroy(pAppendEntry);
}
} else {
// no append entries, do nothing
// maybe has extra entries, no harm
// update match index
pReply->matchIndex = pMsg->prevLogIndex;
}
// maybe update commit index, leader notice me
syncNodeFollowerCommit(ths, pMsg->commitIndex);
syncLogRecvAppendEntries(ths, pMsg, "accept");
goto _SEND_RESPONSE;
_IGNORE:
rpcFreeCont(rpcRsp.pCont);
return 0;
_SEND_RESPONSE:
// msg event log
syncLogSendAppendEntriesReply(ths, pReply, "");
// send response
syncNodeSendMsgById(&pReply->destId, ths, &rpcRsp);
return 0;
}
source/libs/sync/src/syncAppendEntriesReply.c
浏览文件 @ 0d2d3349
......@@ -89,63 +89,3 @@ int32_t syncNodeOnAppendEntriesReply(SSyncNode* ths, const SRpcMsg* pRpcMsg) {
}
return 0;
}
int32_t syncNodeOnAppendEntriesReplyOld(SSyncNode* ths, SyncAppendEntriesReply* pMsg) {
int32_t ret = 0;
// if already drop replica, do not process
if (!syncNodeInRaftGroup(ths, &(pMsg->srcId))) {
syncLogRecvAppendEntriesReply(ths, pMsg, "not in my config");
return 0;
}
// drop stale response
if (pMsg->term < ths->raftStore.currentTerm) {
syncLogRecvAppendEntriesReply(ths, pMsg, "drop stale response");
return 0;
}
if (ths->state == TAOS_SYNC_STATE_LEADER) {
if (pMsg->term > ths->raftStore.currentTerm) {
syncLogRecvAppendEntriesReply(ths, pMsg, "error term");
syncNodeStepDown(ths, pMsg->term);
return -1;
}
ASSERT(pMsg->term == ths->raftStore.currentTerm);
if (pMsg->success) {
SyncIndex oldMatchIndex = syncIndexMgrGetIndex(ths->pMatchIndex, &(pMsg->srcId));
if (pMsg->matchIndex > oldMatchIndex) {
syncIndexMgrSetIndex(ths->pMatchIndex, &(pMsg->srcId), pMsg->matchIndex);
syncMaybeAdvanceCommitIndex(ths);
// maybe update minMatchIndex
ths->minMatchIndex = syncMinMatchIndex(ths);
}
syncIndexMgrSetIndex(ths->pNextIndex, &(pMsg->srcId), pMsg->matchIndex + 1);
} else {
SyncIndex nextIndex = syncIndexMgrGetIndex(ths->pNextIndex, &(pMsg->srcId));
if (nextIndex > SYNC_INDEX_BEGIN) {
--nextIndex;
}
syncIndexMgrSetIndex(ths->pNextIndex, &(pMsg->srcId), nextIndex);
}
// send next append entries
SPeerState* pState = syncNodeGetPeerState(ths, &(pMsg->srcId));
ASSERT(pState != NULL);
if (pMsg->lastSendIndex == pState->lastSendIndex) {
int64_t timeNow = taosGetTimestampMs();
int64_t elapsed = timeNow - pState->lastSendTime;
sNTrace(ths, "sync-append-entries rtt elapsed:%" PRId64 ", index:%" PRId64, elapsed, pState->lastSendIndex);
syncNodeReplicateOne(ths, &(pMsg->srcId), true);
}
}
syncLogRecvAppendEntriesReply(ths, pMsg, "process");
return 0;
}
source/libs/sync/src/syncCommit.c
浏览文件 @ 0d2d3349
......@@ -43,148 +43,6 @@
// IN commitIndex' = [commitIndex EXCEPT ![i] = newCommitIndex]
// /\ UNCHANGED <<messages, serverVars, candidateVars, leaderVars, log>>
//
void syncOneReplicaAdvance(SSyncNode* pSyncNode) {
ASSERT(false && "deprecated");
if (pSyncNode == NULL) {
sError("pSyncNode is NULL");
return;
}
if (pSyncNode->state != TAOS_SYNC_STATE_LEADER) {
sNError(pSyncNode, "not leader, can not advance commit index");
return;
}
if (pSyncNode->replicaNum != 1) {
sNError(pSyncNode, "not one replica, can not advance commit index");
return;
}
// advance commit index to snapshot first
SSnapshot snapshot;
pSyncNode->pFsm->FpGetSnapshotInfo(pSyncNode->pFsm, &snapshot);
if (snapshot.lastApplyIndex > 0 && snapshot.lastApplyIndex > pSyncNode->commitIndex) {
SyncIndex commitBegin = pSyncNode->commitIndex;
SyncIndex commitEnd = snapshot.lastApplyIndex;
pSyncNode->commitIndex = snapshot.lastApplyIndex;
sNTrace(pSyncNode, "commit by snapshot from index:%" PRId64 " to index:%" PRId64, commitBegin, commitEnd);
}
// advance commit index as large as possible
SyncIndex lastIndex = syncNodeGetLastIndex(pSyncNode);
if (lastIndex > pSyncNode->commitIndex) {
sNTrace(pSyncNode, "commit by wal from index:%" PRId64 " to index:%" PRId64, pSyncNode->commitIndex + 1, lastIndex);
pSyncNode->commitIndex = lastIndex;
}
// call back Wal
SyncIndex walCommitVer = logStoreWalCommitVer(pSyncNode->pLogStore);
if (pSyncNode->commitIndex > walCommitVer) {
pSyncNode->pLogStore->syncLogUpdateCommitIndex(pSyncNode->pLogStore, pSyncNode->commitIndex);
}
}
void syncMaybeAdvanceCommitIndex(SSyncNode* pSyncNode) {
ASSERTS(false, "deprecated");
if (pSyncNode == NULL) {
sError("pSyncNode is NULL");
return;
}
if (pSyncNode->state != TAOS_SYNC_STATE_LEADER) {
sNError(pSyncNode, "not leader, can not advance commit index");
return;
}
// advance commit index to sanpshot first
SSnapshot snapshot;
pSyncNode->pFsm->FpGetSnapshotInfo(pSyncNode->pFsm, &snapshot);
if (snapshot.lastApplyIndex > 0 && snapshot.lastApplyIndex > pSyncNode->commitIndex) {
SyncIndex commitBegin = pSyncNode->commitIndex;
SyncIndex commitEnd = snapshot.lastApplyIndex;
pSyncNode->commitIndex = snapshot.lastApplyIndex;
sNTrace(pSyncNode, "commit by snapshot from index:%" PRId64 " to index:%" PRId64, commitBegin, commitEnd);
}
// update commit index
SyncIndex newCommitIndex = pSyncNode->commitIndex;
for (SyncIndex index = syncNodeGetLastIndex(pSyncNode); index > pSyncNode->commitIndex; --index) {
bool agree = syncAgree(pSyncNode, index);
if (agree) {
// term
SSyncRaftEntry* pEntry = NULL;
SLRUCache* pCache = pSyncNode->pLogStore->pCache;
LRUHandle* h = taosLRUCacheLookup(pCache, &index, sizeof(index));
if (h) {
pEntry = (SSyncRaftEntry*)taosLRUCacheValue(pCache, h);
pSyncNode->pLogStore->cacheHit++;
sNTrace(pSyncNode, "hit cache index:%" PRId64 ", bytes:%u, %p", index, pEntry->bytes, pEntry);
} else {
pSyncNode->pLogStore->cacheMiss++;
sNTrace(pSyncNode, "miss cache index:%" PRId64, index);
int32_t code = pSyncNode->pLogStore->syncLogGetEntry(pSyncNode->pLogStore, index, &pEntry);
if (code != 0) {
sNError(pSyncNode, "advance commit index error, read wal index:%" PRId64, index);
return;
}
}
// cannot commit, even if quorum agree. need check term!
if (pEntry->term <= pSyncNode->raftStore.currentTerm) {
// update commit index
newCommitIndex = index;
if (h) {
taosLRUCacheRelease(pCache, h, false);
} else {
syncEntryDestroy(pEntry);
}
break;
} else {
sNTrace(pSyncNode, "can not commit due to term not equal, index:%" PRId64 ", term:%" PRIu64, pEntry->index,
pEntry->term);
}
if (h) {
taosLRUCacheRelease(pCache, h, false);
} else {
syncEntryDestroy(pEntry);
}
}
}
// advance commit index as large as possible
SyncIndex walCommitVer = logStoreWalCommitVer(pSyncNode->pLogStore);
if (walCommitVer > newCommitIndex) {
newCommitIndex = walCommitVer;
}
// maybe execute fsm
if (newCommitIndex > pSyncNode->commitIndex) {
SyncIndex beginIndex = pSyncNode->commitIndex + 1;
SyncIndex endIndex = newCommitIndex;
// update commit index
pSyncNode->commitIndex = newCommitIndex;
// call back Wal
pSyncNode->pLogStore->syncLogUpdateCommitIndex(pSyncNode->pLogStore, pSyncNode->commitIndex);
// execute fsm
if (pSyncNode != NULL && pSyncNode->pFsm != NULL) {
int32_t code = syncNodeDoCommit(pSyncNode, beginIndex, endIndex, pSyncNode->state);
if (code != 0) {
sNError(pSyncNode, "advance commit index error, do commit begin:%" PRId64 ", end:%" PRId64, beginIndex,
endIndex);
return;
}
}
}
}
bool syncAgreeIndex(SSyncNode* pSyncNode, SRaftId* pRaftId, SyncIndex index) {
// I am leader, I agree
......@@ -210,83 +68,7 @@ static inline int64_t syncNodeAbs64(int64_t a, int64_t b) {
return c;
}
int32_t syncNodeDynamicQuorum(const SSyncNode* pSyncNode) {
return pSyncNode->quorum;
#if 0
int32_t quorum = 1; // self
int64_t timeNow = taosGetTimestampMs();
for (int i = 0; i < pSyncNode->peersNum; ++i) {
int64_t peerStartTime = syncIndexMgrGetStartTime(pSyncNode->pNextIndex, &(pSyncNode->peersId)[i]);
int64_t peerRecvTime = syncIndexMgrGetRecvTime(pSyncNode->pNextIndex, &(pSyncNode->peersId)[i]);
SyncIndex peerMatchIndex = syncIndexMgrGetIndex(pSyncNode->pMatchIndex, &(pSyncNode->peersId)[i]);
int64_t recvTimeDiff = TABS(peerRecvTime - timeNow);
int64_t startTimeDiff = TABS(peerStartTime - pSyncNode->startTime);
int64_t logDiff = TABS(peerMatchIndex - syncNodeGetLastIndex(pSyncNode));
/*
int64_t recvTimeDiff = syncNodeAbs64(peerRecvTime, timeNow);
int64_t startTimeDiff = syncNodeAbs64(peerStartTime, pSyncNode->startTime);
int64_t logDiff = syncNodeAbs64(peerMatchIndex, syncNodeGetLastIndex(pSyncNode));
*/
int32_t addQuorum = 0;
if (recvTimeDiff < SYNC_MAX_RECV_TIME_RANGE_MS) {
if (startTimeDiff < SYNC_MAX_START_TIME_RANGE_MS) {
addQuorum = 1;
} else {
if (logDiff < SYNC_ADD_QUORUM_COUNT) {
addQuorum = 1;
} else {
addQuorum = 0;
}
}
} else {
addQuorum = 0;
}
/*
if (recvTimeDiff < SYNC_MAX_RECV_TIME_RANGE_MS) {
addQuorum = 1;
} else {
addQuorum = 0;
}
if (startTimeDiff > SYNC_MAX_START_TIME_RANGE_MS) {
addQuorum = 0;
}
*/
quorum += addQuorum;
}
ASSERT(quorum <= pSyncNode->replicaNum);
if (quorum < pSyncNode->quorum) {
quorum = pSyncNode->quorum;
}
return quorum;
#endif
}
/*
bool syncAgree(SSyncNode* pSyncNode, SyncIndex index) {
int agreeCount = 0;
for (int i = 0; i < pSyncNode->replicaNum; ++i) {
if (syncAgreeIndex(pSyncNode, &(pSyncNode->replicasId[i]), index)) {
++agreeCount;
}
if (agreeCount >= syncNodeDynamicQuorum(pSyncNode)) {
return true;
}
}
return false;
}
*/
int32_t syncNodeDynamicQuorum(const SSyncNode* pSyncNode) { return pSyncNode->quorum; }
bool syncNodeAgreedUpon(SSyncNode* pNode, SyncIndex index) {
int count = 0;
......
source/libs/sync/src/syncElection.c
浏览文件 @ 0d2d3349
......@@ -43,7 +43,10 @@ static int32_t syncNodeRequestVotePeers(SSyncNode* pNode) {
for (int i = 0; i < pNode->peersNum; ++i) {
SRpcMsg rpcMsg = {0};
ret = syncBuildRequestVote(&rpcMsg, pNode->vgId);
ASSERT(ret == 0);
if (ret < 0) {
sError("vgId:%d, failed to build request-vote msg since %s", pNode->vgId, terrstr());
continue;
}
SyncRequestVote* pMsg = rpcMsg.pCont;
pMsg->srcId = pNode->myRaftId;
......@@ -51,13 +54,18 @@ static int32_t syncNodeRequestVotePeers(SSyncNode* pNode) {
pMsg->term = pNode->raftStore.currentTerm;
ret = syncNodeGetLastIndexTerm(pNode, &pMsg->lastLogIndex, &pMsg->lastLogTerm);
ASSERT(ret == 0);
if (ret < 0) {
sError("vgId:%d, failed to get index and term of last log since %s", pNode->vgId, terrstr());
continue;
}
ret = syncNodeSendMsgById(&pNode->peersId[i], pNode, &rpcMsg);
ASSERT(ret == 0);
if (ret < 0) {
sError("vgId:%d, failed to send msg to peerId:%" PRId64, pNode->vgId, pNode->peersId[i].addr);
continue;
}
return ret;
}
return 0;
}
int32_t syncNodeElect(SSyncNode* pSyncNode) {
......
source/libs/sync/src/syncMain.c
浏览文件 @ 0d2d3349
......@@ -292,8 +292,6 @@ int32_t syncBeginSnapshot(int64_t rid, int64_t lastApplyIndex) {
goto _DEL_WAL;
} else {
lastApplyIndex -= SYNC_VNODE_LOG_RETENTION;
SyncIndex beginIndex = pSyncNode->pLogStore->syncLogBeginIndex(pSyncNode->pLogStore);
SyncIndex endIndex = pSyncNode->pLogStore->syncLogEndIndex(pSyncNode->pLogStore);
bool isEmpty = pSyncNode->pLogStore->syncLogIsEmpty(pSyncNode->pLogStore);
......@@ -308,6 +306,8 @@ int32_t syncBeginSnapshot(int64_t rid, int64_t lastApplyIndex) {
if (pSyncNode->replicaNum > 1) {
// multi replicas
lastApplyIndex = TMAX(lastApplyIndex - SYNC_VNODE_LOG_RETENTION, beginIndex - 1);
if (pSyncNode->state == TAOS_SYNC_STATE_LEADER) {
pSyncNode->minMatchIndex = syncMinMatchIndex(pSyncNode);
......@@ -586,78 +586,6 @@ SSyncState syncGetState(int64_t rid) {
return state;
}
#if 0
int32_t syncGetSnapshotByIndex(int64_t rid, SyncIndex index, SSnapshot* pSnapshot) {
if (index < SYNC_INDEX_BEGIN) {
return -1;
}
SSyncNode* pSyncNode = syncNodeAcquire(rid);
if (pSyncNode == NULL) {
return -1;
}
ASSERT(rid == pSyncNode->rid);
SSyncRaftEntry* pEntry = NULL;
int32_t code = pSyncNode->pLogStore->syncLogGetEntry(pSyncNode->pLogStore, index, &pEntry);
if (code != 0) {
if (pEntry != NULL) {
syncEntryDestroy(pEntry);
}
syncNodeRelease(pSyncNode);
return -1;
}
ASSERT(pEntry != NULL);
pSnapshot->data = NULL;
pSnapshot->lastApplyIndex = index;
pSnapshot->lastApplyTerm = pEntry->term;
pSnapshot->lastConfigIndex = syncNodeGetSnapshotConfigIndex(pSyncNode, index);
syncEntryDestroy(pEntry);
syncNodeRelease(pSyncNode);
return 0;
}
int32_t syncGetSnapshotMeta(int64_t rid, struct SSnapshotMeta* sMeta) {
SSyncNode* pSyncNode = syncNodeAcquire(rid);
if (pSyncNode == NULL) {
return -1;
}
ASSERT(rid == pSyncNode->rid);
sMeta->lastConfigIndex = pSyncNode->raftCfg.lastConfigIndex;
sTrace("vgId:%d, get snapshot meta, lastConfigIndex:%" PRId64, pSyncNode->vgId, pSyncNode->raftCfg.lastConfigIndex);
syncNodeRelease(pSyncNode);
return 0;
}
int32_t syncGetSnapshotMetaByIndex(int64_t rid, SyncIndex snapshotIndex, struct SSnapshotMeta* sMeta) {
SSyncNode* pSyncNode = syncNodeAcquire(rid);
if (pSyncNode == NULL) {
return -1;
}
ASSERT(rid == pSyncNode->rid);
ASSERT(pSyncNode->raftCfg.configIndexCount >= 1);
SyncIndex lastIndex = (pSyncNode->raftCfg.configIndexArr)[0];
for (int32_t i = 0; i < pSyncNode->raftCfg.configIndexCount; ++i) {
if ((pSyncNode->raftCfg.configIndexArr)[i] > lastIndex &&
(pSyncNode->raftCfg.configIndexArr)[i] <= snapshotIndex) {
lastIndex = (pSyncNode->raftCfg.configIndexArr)[i];
}
}
sMeta->lastConfigIndex = lastIndex;
sTrace("vgId:%d, get snapshot meta by index:%" PRId64 " lcindex:%" PRId64, pSyncNode->vgId, snapshotIndex,
sMeta->lastConfigIndex);
syncNodeRelease(pSyncNode);
return 0;
}
#endif
SyncIndex syncNodeGetSnapshotConfigIndex(SSyncNode* pSyncNode, SyncIndex snapshotLastApplyIndex) {
ASSERT(pSyncNode->raftCfg.configIndexCount >= 1);
SyncIndex lastIndex = (pSyncNode->raftCfg.configIndexArr)[0];
......@@ -1031,9 +959,12 @@ SSyncNode* syncNodeOpen(SSyncInfo* pSyncInfo) {
pSyncNode->commitIndex = commitIndex;
sInfo("vgId:%d, sync node commitIndex initialized as %" PRId64, pSyncNode->vgId, pSyncNode->commitIndex);
// restore log store on need
if (syncNodeLogStoreRestoreOnNeed(pSyncNode) < 0) {
sError("vgId:%d, failed to restore log store since %s.", pSyncNode->vgId, terrstr());
goto _error;
}
// timer ms init
pSyncNode->pingBaseLine = PING_TIMER_MS;
pSyncNode->electBaseLine = tsElectInterval;
......@@ -1096,10 +1027,16 @@ SSyncNode* syncNodeOpen(SSyncInfo* pSyncInfo) {
pSyncNode->changing = false;
// replication mgr
syncNodeLogReplMgrInit(pSyncNode);
if (syncNodeLogReplMgrInit(pSyncNode) < 0) {
sError("vgId:%d, failed to init repl mgr since %s.", pSyncNode->vgId, terrstr());
goto _error;
}
// peer state
syncNodePeerStateInit(pSyncNode);
if (syncNodePeerStateInit(pSyncNode) < 0) {
sError("vgId:%d, failed to init peer stat since %s.", pSyncNode->vgId, terrstr());
goto _error;
}
//
// min match index
......@@ -1194,27 +1131,10 @@ int32_t syncNodeStart(SSyncNode* pSyncNode) {
int32_t ret = 0;
ret = syncNodeStartPingTimer(pSyncNode);
ASSERT(ret == 0);
return ret;
}
void syncNodeStartOld(SSyncNode* pSyncNode) {
// start raft
if (pSyncNode->replicaNum == 1) {
raftStoreNextTerm(pSyncNode);
syncNodeBecomeLeader(pSyncNode, "one replica start");
// Raft 3.6.2 Committing entries from previous terms
syncNodeAppendNoop(pSyncNode);
syncMaybeAdvanceCommitIndex(pSyncNode);
} else {
syncNodeBecomeFollower(pSyncNode, "first start");
if (ret != 0) {
sError("vgId:%d, failed to start ping timer since %s", pSyncNode->vgId, terrstr());
}
int32_t ret = 0;
ret = syncNodeStartPingTimer(pSyncNode);
ASSERT(ret == 0);
return ret;
}
int32_t syncNodeStartStandBy(SSyncNode* pSyncNode) {
......@@ -1225,11 +1145,16 @@ int32_t syncNodeStartStandBy(SSyncNode* pSyncNode) {
// reset elect timer, long enough
int32_t electMS = TIMER_MAX_MS;
int32_t ret = syncNodeRestartElectTimer(pSyncNode, electMS);
ASSERT(ret == 0);
if (ret < 0) {
sError("vgId:%d, failed to restart elect timer since %s", pSyncNode->vgId, terrstr());
return -1;
}
ret = 0;
ret = syncNodeStartPingTimer(pSyncNode);
ASSERT(ret == 0);
if (ret < 0) {
sError("vgId:%d, failed to start ping timer since %s", pSyncNode->vgId, terrstr());
return -1;
}
return ret;
}
......@@ -1818,12 +1743,6 @@ void syncNodeBecomeLeader(SSyncNode* pSyncNode, const char* debugStr) {
pSyncNode->leaderCache = pSyncNode->myRaftId;
for (int32_t i = 0; i < pSyncNode->pNextIndex->replicaNum; ++i) {
// maybe overwrite myself, no harm
// just do it!
// pSyncNode->pNextIndex->index[i] = pSyncNode->pLogStore->getLastIndex(pSyncNode->pLogStore) + 1;
// maybe wal is deleted
SyncIndex lastIndex;
SyncTerm lastTerm;
int32_t code = syncNodeGetLastIndexTerm(pSyncNode, &lastIndex, &lastTerm);
......@@ -1885,7 +1804,11 @@ void syncNodeBecomeLeader(SSyncNode* pSyncNode, const char* debugStr) {
void syncNodeCandidate2Leader(SSyncNode* pSyncNode) {
ASSERT(pSyncNode->state == TAOS_SYNC_STATE_CANDIDATE);
ASSERT(voteGrantedMajority(pSyncNode->pVotesGranted));
bool granted = voteGrantedMajority(pSyncNode->pVotesGranted);
if (!granted) {
sError("vgId:%d, not granted by majority.", pSyncNode->vgId);
return;
}
syncNodeBecomeLeader(pSyncNode, "candidate to leader");
sNTrace(pSyncNode, "state change syncNodeCandidate2Leader");
......@@ -1901,20 +1824,6 @@ void syncNodeCandidate2Leader(SSyncNode* pSyncNode) {
pSyncNode->vgId, pSyncNode->raftStore.currentTerm, pSyncNode->commitIndex, lastIndex);
}
void syncNodeCandidate2LeaderOld(SSyncNode* pSyncNode) {
ASSERT(pSyncNode->state == TAOS_SYNC_STATE_CANDIDATE);
ASSERT(voteGrantedMajority(pSyncNode->pVotesGranted));
syncNodeBecomeLeader(pSyncNode, "candidate to leader");
// Raft 3.6.2 Committing entries from previous terms
syncNodeAppendNoop(pSyncNode);
syncMaybeAdvanceCommitIndex(pSyncNode);
if (pSyncNode->replicaNum > 1) {
syncNodeReplicate(pSyncNode);
}
}
bool syncNodeIsMnode(SSyncNode* pSyncNode) { return (pSyncNode->vgId == 1); }
int32_t syncNodePeerStateInit(SSyncNode* pSyncNode) {
......@@ -1960,7 +1869,8 @@ void syncNodeCandidate2Follower(SSyncNode* pSyncNode) {
// need assert
void syncNodeVoteForTerm(SSyncNode* pSyncNode, SyncTerm term, SRaftId* pRaftId) {
ASSERT(term == pSyncNode->raftStore.currentTerm);
ASSERT(!raftStoreHasVoted(pSyncNode));
bool voted = raftStoreHasVoted(pSyncNode);
ASSERT(!voted);
raftStoreVote(pSyncNode, pRaftId);
}
......@@ -2638,24 +2548,6 @@ int32_t syncNodeOnLocalCmd(SSyncNode* ths, const SRpcMsg* pRpcMsg) {
return 0;
}
int32_t syncNodeOnLocalCmdOld(SSyncNode* ths, const SRpcMsg* pRpcMsg) {
ASSERT(false && "deprecated");
SyncLocalCmd* pMsg = pRpcMsg->pCont;
syncLogRecvLocalCmd(ths, pMsg, "");
if (pMsg->cmd == SYNC_LOCAL_CMD_STEP_DOWN) {
syncNodeStepDown(ths, pMsg->currentTerm);
} else if (pMsg->cmd == SYNC_LOCAL_CMD_FOLLOWER_CMT) {
syncNodeFollowerCommit(ths, pMsg->commitIndex);
} else {
sError("error local cmd");
}
return 0;
}
// TLA+ Spec
// ClientRequest(i, v) ==
// /\ state[i] = Leader
......@@ -2700,96 +2592,6 @@ int32_t syncNodeOnClientRequest(SSyncNode* ths, SRpcMsg* pMsg, SyncIndex* pRetIn
}
}
int32_t syncNodeOnClientRequestOld(SSyncNode* ths, SRpcMsg* pMsg, SyncIndex* pRetIndex) {
sNTrace(ths, "on client request");
int32_t ret = 0;
int32_t code = 0;
SyncIndex index = ths->pLogStore->syncLogWriteIndex(ths->pLogStore);
SyncTerm term = ths->raftStore.currentTerm;
SSyncRaftEntry* pEntry;
if (pMsg->msgType == TDMT_SYNC_CLIENT_REQUEST) {
pEntry = syncEntryBuildFromClientRequest(pMsg->pCont, term, index);
} else {
pEntry = syncEntryBuildFromRpcMsg(pMsg, term, index);
}
LRUHandle* h = NULL;
if (ths->state == TAOS_SYNC_STATE_LEADER) {
// append entry
code = ths->pLogStore->syncLogAppendEntry(ths->pLogStore, pEntry, false);
if (code != 0) {
if (ths->replicaNum == 1) {
if (h) {
taosLRUCacheRelease(ths->pLogStore->pCache, h, false);
} else {
syncEntryDestroy(pEntry);
}
return -1;
} else {
// del resp mgr, call FpCommitCb
SFsmCbMeta cbMeta = {
.index = pEntry->index,
.lastConfigIndex = SYNC_INDEX_INVALID,
.isWeak = pEntry->isWeak,
.code = -1,
.state = ths->state,
.seqNum = pEntry->seqNum,
.term = pEntry->term,
.currentTerm = ths->raftStore.currentTerm,
.flag = 0,
};
ths->pFsm->FpCommitCb(ths->pFsm, pMsg, &cbMeta);
if (h) {
taosLRUCacheRelease(ths->pLogStore->pCache, h, false);
} else {
syncEntryDestroy(pEntry);
}
return -1;
}
}
syncCacheEntry(ths->pLogStore, pEntry, &h);
// if mulit replica, start replicate right now
if (ths->replicaNum > 1) {
syncNodeReplicate(ths);
}
// if only myself, maybe commit right now
if (ths->replicaNum == 1) {
if (syncNodeIsMnode(ths)) {
syncMaybeAdvanceCommitIndex(ths);
} else {
syncOneReplicaAdvance(ths);
}
}
}
if (pRetIndex != NULL) {
if (ret == 0 && pEntry != NULL) {
*pRetIndex = pEntry->index;
} else {
*pRetIndex = SYNC_INDEX_INVALID;
}
}
if (h) {
taosLRUCacheRelease(ths->pLogStore->pCache, h, false);
} else {
syncEntryDestroy(pEntry);
}
return ret;
}
const char* syncStr(ESyncState state) {
switch (state) {
case TAOS_SYNC_STATE_FOLLOWER:
......@@ -2894,129 +2696,6 @@ bool syncNodeIsOptimizedOneReplica(SSyncNode* ths, SRpcMsg* pMsg) {
return (ths->replicaNum == 1 && syncUtilUserCommit(pMsg->msgType) && ths->vgId != 1);
}
int32_t syncNodeDoCommit(SSyncNode* ths, SyncIndex beginIndex, SyncIndex endIndex, uint64_t flag) {
ASSERT(false);
if (beginIndex > endIndex) {
return 0;
}
if (ths == NULL) {
return -1;
}
if (ths->pFsm != NULL && ths->pFsm->FpGetSnapshotInfo != NULL) {
// advance commit index to sanpshot first
SSnapshot snapshot = {0};
ths->pFsm->FpGetSnapshotInfo(ths->pFsm, &snapshot);
if (snapshot.lastApplyIndex >= 0 && snapshot.lastApplyIndex >= beginIndex) {
sNTrace(ths, "commit by snapshot from index:%" PRId64 " to index:%" PRId64, beginIndex, snapshot.lastApplyIndex);
// update begin index
beginIndex = snapshot.lastApplyIndex + 1;
}
}
int32_t code = 0;
ESyncState state = flag;
sNTrace(ths, "commit by wal from index:%" PRId64 " to index:%" PRId64, beginIndex, endIndex);
// execute fsm
if (ths->pFsm != NULL) {
for (SyncIndex i = beginIndex; i <= endIndex; ++i) {
if (i != SYNC_INDEX_INVALID) {
SSyncRaftEntry* pEntry;
SLRUCache* pCache = ths->pLogStore->pCache;
LRUHandle* h = taosLRUCacheLookup(pCache, &i, sizeof(i));
if (h) {
pEntry = (SSyncRaftEntry*)taosLRUCacheValue(pCache, h);
ths->pLogStore->cacheHit++;
sNTrace(ths, "hit cache index:%" PRId64 ", bytes:%u, %p", i, pEntry->bytes, pEntry);
} else {
ths->pLogStore->cacheMiss++;
sNTrace(ths, "miss cache index:%" PRId64, i);
code = ths->pLogStore->syncLogGetEntry(ths->pLogStore, i, &pEntry);
// ASSERT(code == 0);
// ASSERT(pEntry != NULL);
if (code != 0 || pEntry == NULL) {
sNError(ths, "get log entry error");
sFatal("vgId:%d, get log entry %" PRId64 " error when commit since %s", ths->vgId, i, terrstr());
continue;
}
}
SRpcMsg rpcMsg = {0};
syncEntry2OriginalRpc(pEntry, &rpcMsg);
sTrace("do commit index:%" PRId64 ", type:%s", i, TMSG_INFO(pEntry->msgType));
// user commit
if ((ths->pFsm->FpCommitCb != NULL) && syncUtilUserCommit(pEntry->originalRpcType)) {
bool internalExecute = true;
if ((ths->replicaNum == 1) && ths->restoreFinish && ths->vgId != 1) {
internalExecute = false;
}
sNTrace(ths, "user commit index:%" PRId64 ", internal:%d, type:%s", i, internalExecute,
TMSG_INFO(pEntry->msgType));
// execute fsm in apply thread, or execute outside syncPropose
if (internalExecute) {
SFsmCbMeta cbMeta = {
.index = pEntry->index,
.lastConfigIndex = syncNodeGetSnapshotConfigIndex(ths, pEntry->index),
.isWeak = pEntry->isWeak,
.code = 0,
.state = ths->state,
.seqNum = pEntry->seqNum,
.term = pEntry->term,
.currentTerm = ths->raftStore.currentTerm,
.flag = flag,
};
syncRespMgrGetAndDel(ths->pSyncRespMgr, cbMeta.seqNum, &rpcMsg.info);
ths->pFsm->FpCommitCb(ths->pFsm, &rpcMsg, &cbMeta);
}
}
#if 0
// execute in pre-commit
// leader transfer
if (pEntry->originalRpcType == TDMT_SYNC_LEADER_TRANSFER) {
code = syncDoLeaderTransfer(ths, &rpcMsg, pEntry);
ASSERT(code == 0);
}
#endif
// restore finish
// if only snapshot, a noop entry will be append, so syncLogLastIndex is always ok
if (pEntry->index == ths->pLogStore->syncLogLastIndex(ths->pLogStore)) {
if (ths->restoreFinish == false) {
if (ths->pFsm->FpRestoreFinishCb != NULL) {
ths->pFsm->FpRestoreFinishCb(ths->pFsm);
}
ths->restoreFinish = true;
int64_t restoreDelay = taosGetTimestampMs() - ths->leaderTime;
sNTrace(ths, "restore finish, index:%" PRId64 ", elapsed:%" PRId64 " ms", pEntry->index, restoreDelay);
}
}
rpcFreeCont(rpcMsg.pCont);
if (h) {
taosLRUCacheRelease(pCache, h, false);
} else {
syncEntryDestroy(pEntry);
}
}
}
}
return 0;
}
bool syncNodeInRaftGroup(SSyncNode* ths, SRaftId* pRaftId) {
for (int32_t i = 0; i < ths->replicaNum; ++i) {
if (syncUtilSameId(&((ths->replicasId)[i]), pRaftId)) {
......
source/libs/sync/src/syncPipeline.c
浏览文件 @ 0d2d3349
......@@ -945,8 +945,11 @@ int32_t syncNodeLogReplMgrInit(SSyncNode* pNode) {
for (int i = 0; i < TSDB_MAX_REPLICA; i++) {
ASSERT(pNode->logReplMgrs[i] == NULL);
pNode->logReplMgrs[i] = syncLogReplMgrCreate();
if (pNode->logReplMgrs[i] == NULL) {
terrno = TSDB_CODE_OUT_OF_MEMORY;
return -1;
}
pNode->logReplMgrs[i]->peerId = i;
ASSERTS(pNode->logReplMgrs[i] != NULL, "Out of memory.");
}
return 0;
}
......
source/libs/sync/src/syncReplication.c
浏览文件 @ 0d2d3349
......@@ -48,92 +48,6 @@
int32_t syncNodeMaybeSendAppendEntries(SSyncNode* pSyncNode, const SRaftId* destRaftId, SRpcMsg* pRpcMsg);
int32_t syncNodeReplicateOne(SSyncNode* pSyncNode, SRaftId* pDestId, bool snapshot) {
ASSERT(false && "deprecated");
// next index
SyncIndex nextIndex = syncIndexMgrGetIndex(pSyncNode->pNextIndex, pDestId);
if (snapshot) {
// maybe start snapshot
SyncIndex logStartIndex = pSyncNode->pLogStore->syncLogBeginIndex(pSyncNode->pLogStore);
SyncIndex logEndIndex = pSyncNode->pLogStore->syncLogEndIndex(pSyncNode->pLogStore);
if (nextIndex < logStartIndex || nextIndex - 1 > logEndIndex) {
sNTrace(pSyncNode, "maybe start snapshot for next-index:%" PRId64 ", start:%" PRId64 ", end:%" PRId64, nextIndex,
logStartIndex, logEndIndex);
// start snapshot
int32_t code = syncNodeStartSnapshot(pSyncNode, pDestId);
}
}
// pre index, pre term
SyncIndex preLogIndex = syncNodeGetPreIndex(pSyncNode, nextIndex);
SyncTerm preLogTerm = syncNodeGetPreTerm(pSyncNode, nextIndex);
// prepare entry
SRpcMsg rpcMsg = {0};
SyncAppendEntries* pMsg = NULL;
SSyncRaftEntry* pEntry = NULL;
SLRUCache* pCache = pSyncNode->pLogStore->pCache;
LRUHandle* h = taosLRUCacheLookup(pCache, &nextIndex, sizeof(nextIndex));
int32_t code = 0;
if (h) {
pEntry = (SSyncRaftEntry*)taosLRUCacheValue(pCache, h);
code = 0;
pSyncNode->pLogStore->cacheHit++;
sNTrace(pSyncNode, "hit cache index:%" PRId64 ", bytes:%u, %p", nextIndex, pEntry->bytes, pEntry);
} else {
pSyncNode->pLogStore->cacheMiss++;
sNTrace(pSyncNode, "miss cache index:%" PRId64, nextIndex);
code = pSyncNode->pLogStore->syncLogGetEntry(pSyncNode->pLogStore, nextIndex, &pEntry);
}
if (code == 0) {
ASSERT(pEntry != NULL);
code = syncBuildAppendEntries(&rpcMsg, (int32_t)(pEntry->bytes), pSyncNode->vgId);
ASSERT(code == 0);
pMsg = rpcMsg.pCont;
memcpy(pMsg->data, pEntry, pEntry->bytes);
} else {
if (terrno == TSDB_CODE_WAL_LOG_NOT_EXIST) {
// no entry in log
code = syncBuildAppendEntries(&rpcMsg, 0, pSyncNode->vgId);
ASSERT(code == 0);
pMsg = rpcMsg.pCont;
} else {
sNError(pSyncNode, "replicate to dnode:%d error, next-index:%" PRId64, DID(pDestId), nextIndex);
return -1;
}
}
if (h) {
taosLRUCacheRelease(pCache, h, false);
} else {
syncEntryDestroy(pEntry);
}
// prepare msg
ASSERT(pMsg != NULL);
pMsg->srcId = pSyncNode->myRaftId;
pMsg->destId = *pDestId;
pMsg->term = pSyncNode->raftStore.currentTerm;
pMsg->prevLogIndex = preLogIndex;
pMsg->prevLogTerm = preLogTerm;
pMsg->commitIndex = pSyncNode->commitIndex;
pMsg->privateTerm = 0;
// pMsg->privateTerm = syncIndexMgrGetTerm(pSyncNode->pNextIndex, pDestId);
// send msg
syncNodeMaybeSendAppendEntries(pSyncNode, pDestId, &rpcMsg);
return 0;
}
int32_t syncNodeReplicate(SSyncNode* pNode) {
SSyncLogBuffer* pBuf = pNode->pLogBuf;
taosThreadMutexLock(&pBuf->mutex);
......@@ -156,25 +70,6 @@ int32_t syncNodeReplicateWithoutLock(SSyncNode* pNode) {
return 0;
}
int32_t syncNodeReplicateOld(SSyncNode* pSyncNode) {
if (pSyncNode->state != TAOS_SYNC_STATE_LEADER) {
return -1;
}
sNTrace(pSyncNode, "do replicate");
int32_t ret = 0;
for (int i = 0; i < pSyncNode->peersNum; ++i) {
SRaftId* pDestId = &(pSyncNode->peersId[i]);
ret = syncNodeReplicateOne(pSyncNode, pDestId, true);
if (ret != 0) {
sError("vgId:%d, do append entries error for dnode:%d", pSyncNode->vgId, DID(pDestId));
}
}
return 0;
}
int32_t syncNodeSendAppendEntries(SSyncNode* pSyncNode, const SRaftId* destRaftId, SRpcMsg* pRpcMsg) {
SyncAppendEntries* pMsg = pRpcMsg->pCont;
pMsg->destId = *destRaftId;
......@@ -182,39 +77,6 @@ int32_t syncNodeSendAppendEntries(SSyncNode* pSyncNode, const SRaftId* destRaftI
return 0;
}
int32_t syncNodeSendAppendEntriesOld(SSyncNode* pSyncNode, const SRaftId* destRaftId, SRpcMsg* pRpcMsg) {
int32_t ret = 0;
SyncAppendEntries* pMsg = pRpcMsg->pCont;
if (pMsg == NULL) {
sError("vgId:%d, sync-append-entries msg is NULL", pSyncNode->vgId);
return 0;
}
SPeerState* pState = syncNodeGetPeerState(pSyncNode, destRaftId);
if (pState == NULL) {
sError("vgId:%d, replica maybe dropped", pSyncNode->vgId);
return 0;
}
// save index, otherwise pMsg will be free by rpc
SyncIndex saveLastSendIndex = pState->lastSendIndex;
bool update = false;
if (pMsg->dataLen > 0) {
saveLastSendIndex = pMsg->prevLogIndex + 1;
update = true;
}
syncLogSendAppendEntries(pSyncNode, pMsg, "");
syncNodeSendMsgById(destRaftId, pSyncNode, pRpcMsg);
if (update) {
pState->lastSendIndex = saveLastSendIndex;
pState->lastSendTime = taosGetTimestampMs();
}
return ret;
}
int32_t syncNodeMaybeSendAppendEntries(SSyncNode* pSyncNode, const SRaftId* destRaftId, SRpcMsg* pRpcMsg) {
int32_t ret = 0;
SyncAppendEntries* pMsg = pRpcMsg->pCont;
......
source/libs/wal/src/walMeta.c
浏览文件 @ 0d2d3349
......@@ -325,6 +325,35 @@ bool walLogEntriesComplete(const SWal* pWal) {
return complete;
}
int walTrimIdxFile(SWal* pWal, int32_t fileIdx) {
SWalFileInfo* pFileInfo = taosArrayGet(pWal->fileInfoSet, fileIdx);
ASSERT(pFileInfo != NULL);
char fnameStr[WAL_FILE_LEN];
walBuildIdxName(pWal, pFileInfo->firstVer, fnameStr);
int64_t fileSize = 0;
taosStatFile(fnameStr, &fileSize, NULL);
int64_t records = TMAX(0, pFileInfo->lastVer - pFileInfo->firstVer + 1);
int64_t lastEndOffset = records * sizeof(SWalIdxEntry);
if (fileSize <= lastEndOffset) {
return 0;
}
TdFilePtr pFile = taosOpenFile(fnameStr, TD_FILE_READ | TD_FILE_WRITE);
if (pFile == NULL) {
terrno = TAOS_SYSTEM_ERROR(errno);
return -1;
}
wInfo("vgId:%d, trim idx file. file: %s, size: %" PRId64 ", offset: %" PRId64, pWal->cfg.vgId, fnameStr, fileSize,
lastEndOffset);
taosFtruncateFile(pFile, lastEndOffset);
taosCloseFile(&pFile);
return 0;
}
int walCheckAndRepairMeta(SWal* pWal) {
// load log files, get first/snapshot/last version info
const char* logPattern = "^[0-9]+.log$";
......@@ -402,6 +431,8 @@ int walCheckAndRepairMeta(SWal* pWal) {
}
updateMeta = true;
(void)walTrimIdxFile(pWal, fileIdx);
int64_t lastVer = walScanLogGetLastVer(pWal, fileIdx);
if (lastVer < 0) {
if (terrno != TSDB_CODE_WAL_LOG_NOT_EXIST) {
......@@ -567,6 +598,7 @@ int walCheckAndRepairIdxFile(SWal* pWal, int32_t fileIdx) {
goto _err;
}
int64_t count = 0;
while (idxEntry.ver < pFileInfo->lastVer) {
ASSERT(idxEntry.ver == ckHead.head.version);
......@@ -578,11 +610,11 @@ int walCheckAndRepairIdxFile(SWal* pWal, int32_t fileIdx) {
idxEntry.offset, fLogNameStr);
goto _err;
}
wWarn("vgId:%d, wal idx append new entry %" PRId64 " %" PRId64, pWal->cfg.vgId, idxEntry.ver, idxEntry.offset);
if (taosWriteFile(pIdxFile, &idxEntry, sizeof(SWalIdxEntry)) < 0) {
wError("vgId:%d, failed to append file since %s. file:%s", pWal->cfg.vgId, terrstr(), fnameStr);
goto _err;
}
count++;
}
if (taosFsyncFile(pIdxFile) < 0) {
......@@ -590,6 +622,11 @@ int walCheckAndRepairIdxFile(SWal* pWal, int32_t fileIdx) {
goto _err;
}
if (count > 0) {
wInfo("vgId:%d, rebuilt %" PRId64 " wal idx entries until lastVer: %" PRId64, pWal->cfg.vgId, count,
pFileInfo->lastVer);
}
(void)taosCloseFile(&pLogFile);
(void)taosCloseFile(&pIdxFile);
return 0;
......
source/libs/wal/src/walRef.c
浏览文件 @ 0d2d3349
......@@ -77,6 +77,31 @@ void walUnrefVer(SWalRef *pRef) {
}
#endif
SWalRef *walRefFirstVer(SWal *pWal, SWalRef *pRef) {
if (pRef == NULL) {
pRef = walOpenRef(pWal);
if (pRef == NULL) {
return NULL;
}
}
taosThreadMutexLock(&pWal->mutex);
int64_t ver = walGetFirstVer(pWal);
wDebug("vgId:%d, wal ref version %" PRId64 " for first", pWal->cfg.vgId, ver);
pRef->refVer = ver;
// bsearch in fileSet
SWalFileInfo tmpInfo;
tmpInfo.firstVer = ver;
SWalFileInfo *pRet = taosArraySearch(pWal->fileInfoSet, &tmpInfo, compareWalFileInfo, TD_LE);
ASSERT(pRet != NULL);
pRef->refFile = pRet->firstVer;
taosThreadMutexUnlock(&pWal->mutex);
return pRef;
}
SWalRef *walRefCommittedVer(SWal *pWal) {
SWalRef *pRef = walOpenRef(pWal);
if (pRef == NULL) {
......@@ -86,6 +111,8 @@ SWalRef *walRefCommittedVer(SWal *pWal) {
int64_t ver = walGetCommittedVer(pWal);
wDebug("vgId:%d, wal ref version %" PRId64 " for committed", pWal->cfg.vgId, ver);
pRef->refVer = ver;
// bsearch in fileSet
SWalFileInfo tmpInfo;
......
source/os/src/osMath.c
浏览文件 @ 0d2d3349
......@@ -32,7 +32,18 @@ void swapStr(char* j, char* J, int width) {
}
#endif
int qsortHelper(const void* p1, const void* p2, const void* param) {
__compar_fn_t comparFn = param;
return comparFn(p1, p2);
}
// todo refactor: 1) move away; 2) use merge sort instead; 3) qsort is not a stable sort actually.
void taosSort(void* arr, int64_t sz, int64_t width, __compar_fn_t compar) {
qsort(arr, sz, width, compar);
void taosSort(void* base, int64_t sz, int64_t width, __compar_fn_t compar) {
#ifdef _ALPINE
void* param = compar;
taosqsort(base, width, sz, param, qsortHelper);
#else
qsort(base, sz, width, compar);
#endif
}
source/os/src/osTime.c
浏览文件 @ 0d2d3349
此差异已折叠。
source/util/src/talgo.c
浏览文件 @ 0d2d3349
......@@ -41,12 +41,6 @@ static void median(void *src, int64_t size, int64_t s, int64_t e, const void *pa
ASSERT(comparFn(elePtrAt(src, size, mid), elePtrAt(src, size, s), param) <= 0 &&
comparFn(elePtrAt(src, size, s), elePtrAt(src, size, e), param) <= 0);
#ifdef _DEBUG_VIEW
// tTagsPrints(src[s], pOrderDesc->pColumnModel, &pOrderDesc->orderIdx);
// tTagsPrints(src[mid], pOrderDesc->pColumnModel, &pOrderDesc->orderIdx);
// tTagsPrints(src[e], pOrderDesc->pColumnModel, &pOrderDesc->orderIdx);
#endif
}
static void tInsertSort(void *src, int64_t size, int32_t s, int32_t e, const void *param, __ext_compar_fn_t comparFn,
......@@ -278,14 +272,4 @@ void taosheapsort(void *base, int32_t size, int32_t len, const void *parcompar,
}
taosMemoryFree(buf);
/*
char *buf = taosMemoryCalloc(1, size);
for (i = len - 1; i > 0; i--) {
doswap(elePtrAt(base, size, 0), elePtrAt(base, size, i));
taosheapadjust(base, size, 0, i - 1, parcompar, compar, parswap, swap, maxroot);
}
taosMemoryFreeClear(buf);
*/
}
tests/docs-examples-test/python.sh
浏览文件 @ 0d2d3349
此差异已折叠。
tests/script/tsim/db/alter_replica_13.sim
浏览文件 @ 0d2d3349
......@@ -79,6 +79,7 @@ sql insert into db.ctb6 values(now, 6, "6")
sql insert into db.ctb7 values(now, 7, "7")
sql insert into db.ctb8 values(now, 8, "8")
sql insert into db.ctb9 values(now, 9, "9")
sql flush database db;
print =============== step3: create dnodes
sql create dnode $hostname port 7300
......
tests/script/tsim/stream/basic1.sim
浏览文件 @ 0d2d3349
此差异已折叠。
Markdown is supported
0% 或 .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑！
想要评论请 注册 或