If an load task has a relatively short timeout, then we need to ensure that
each RPC of this task does not get blocked for a long time.
And an RPC is usually blocked for two reasons.
1. handling "memory exceeds limit" in the RPC
If the system finds that the memory occupied by the load exceeds the threshold,
it will select the load channel that occupies the most memory and flush the memtable in it.
this operation is done in the RPC, which may be more time consuming.
2. close the load channel
When the load channel receives the last batch, it will end the task.
It will wait for all memtables flushes to finish synchronously. This process is also time consuming.
Therefore, this PR solves this problem by.
1. Use timeout to determine whether it is a high-priority load task
If the timeout of an load task is relatively short, then we mark it as a high-priority task.
2. not processing "memory exceeds limit" for high priority tasks
3. use a separate flush thread to flush memtable for high priority tasks.
Added bprc stub cache check and reset api, used to test whether the bprc stub cache is available, and reset the bprc stub cache
add a config used for auto check and reset bprc stub
1 Avoid sending tasks if there is no data to be consumed
By fetching latest offset of partition before sending tasks.(Fix [Optimize] Avoid too many abort task in routine load job #6803 )
2 Add a preCheckNeedSchedule phase in update() of routine load.
To avoid taking write lock of job for long time when getting all kafka partitions from kafka server.
3 Upgrade librdkafka's version to 1.7.0 to fix a bug of "Local: Unknown partition"
See offsetsForTimes fails with 'Local: Unknown partition' edenhill/librdkafka#3295
4 Avoid unnecessary storage migration task if there is no that storage medium on BE.
Fix [Bug] Too many unnecessary storage migration tasks #6804
1. This bug is introduced from #6582
2. Optimize the error log of Address used used error msg.
3. Add some document about compilation.
1. Add a custom thirdparty download url.
2. Add a custom com.alibaba maven jar package for DataX.
4. Fix bug that BE crash when closing scan node, introduced from #6622.
1. Fix a memory leak in `collect_iterator.cpp` (Fix#6700)
2. Add a new BE config `max_segment_num_per_rowset` to limit the num of segment in new rowset.(Fix#6701)
3. Make the error msg of stream load more friendly.
1. Add license/total line/release badegs.
2. Add monthly active contributor and contributor growth graph
3. fix a pom.xml bug
4. Modify some routine load log on BE side
Use `commitAsync` to commit offset to kafka, instead of using `commitSync`, which may block for a long time.
Also assign a group.id to routine load if user not specified "property.group.id" property, so that all consumer of
this job will use same group.id instead of a random id for each consume task.
Support when creating a kafka routine load, start consumption from a specified point in time instead of a specific offset.
eg:
```
FROM KAFKA
(
"kafka_broker_list" = "broker1:9092,broker2:9092",
"kafka_topic" = "my_topic",
"property.kafka_default_offsets" = "2021-10-10 11:00:00"
);
or
FROM KAFKA
(
"kafka_broker_list" = "broker1:9092,broker2:9092",
"kafka_topic" = "my_topic",
"kafka_partitions" = "0,1,2",
"kafka_offsets" = "2021-10-10 11:00:00, 2021-10-10 11:00:00, 2021-10-10 12:00:00"
);
```
This PR also reconstructed the analysis method of properties when creating or altering
routine load jobs, and unified the analysis process in the `RoutineLoadDataSourceProperties` class.
Record finished stream load job (both successful job and failed job) into audit log
so that we can see when the stream load job was executed and check the details of stream load jobs.
At present, the application of vlog in the code is quite confusing.
It is inherited from impala VLOG_XX format, and there is also VLOG(number) format.
VLOG(number) format does not have a unified specification, so this pr standardizes the use of VLOG
At present, when some rpc errors occur, the client cannot obtain the error information well.
And this CL change the RPC error returned to client like this:
```
ERROR 1064 (HY000): errCode = 2, detailMessage = there is no scanNode Backend. [10002: in black list(A error
occurred: errorCode=2001 errorMessage:Channel inactive error!)]
ERROR 1064 (HY000): failed to send brpc batch, error=The server is overcrowded, error_text=[E1011]The server is
overcrowded @xx.xx.xx.xx:8060 [R1][E1011]The server is overcrowded @xx.xx.xx.xx:8060 [R2][E1011]The server is
overcrowded @xx.xx.xx.xx:8060 [R3][E1011]The server is overcrowded @xx.xx.xx.xx:8060, client: yy.yy.yy.yy
```
BE can not graceful exit because some threads are running in endless
loop. This patch do the following optimization:
- Use the well encapsulated Thread and ThreadPool instead of std::thread
and std::vector<std::thread>
- Use CountDownLatch in thread's loop condition to avoid endless loop
- Introduce a new class Daemon for daemon works, like tcmalloc_gc,
memory_maintenance and calculate_metrics
- Decouple statistics type TaskWorkerPool and StorageEngine notification
by submit tasks to TaskWorkerPool's queue
- Reorder objects' stop and deconstruct in main(), i.e. stop network
services at first, then internal services
- Use libevent in pthreads mode, by calling evthread_use_pthreads(),
then EvHttpServer can exit gracefully in multi-threads
- Call brpc::Server's Stop() and ClearServices() explicitly
Redesign metrics to 3 layers:
MetricRegistry - MetricEntity - Metrics
MetricRegistry : the register center
MetricEntity : the entity registered on MetricRegistry. Generally a MetricRegistry can be registered on several
MetricEntities, each of MetricEntity is an independent entity, such as server, disk_devices, data_directories, thrift
clients and servers, and so on.
Metric : metrics of an entity. Such as fragment_requests_total on server entity, disk_bytes_read on a disk_device entity,
thrift_opened_clients on a thrift_client entity.
MetricPrototype: the type of a metric. MetricPrototype is a global variable, can be shared by the same metrics across
different MetricEntities.
We can observe the workload of BE, and also it's a way to check
whether there is any problem in BE, like some container increase
too large and lead to OOM.
This patch add the following metrics:
```
Name Description
rowset_count_generated_and_in_use The total count of rowset id generated and in use since BE last start
unused_rowsets_count The total count of unused rowset waiting to be GC
broker_count The total count of brokers in management
data_stream_receiver_count The total count of data stream receivers in management
fragment_endpoint_count The total count of fragment endpoints of data stream in management, should always equal to data_stream_receiver_count
active_scan_context_count The total count of active scan contexts
plan_fragment_count The total count of plan fragments in executing
load_channel_count The total count of load channels in management
result_buffer_block_count The total count of result buffer blocks for queries, each block has a limited queue size (default 1024)
result_block_queue_count The total count of queues for fragments, each queue has a limited size (default 20, by config::max_memory_sink_batch_count)
routine_load_task_count The total count of routine load tasks in executing
small_file_cache_count The total count of cached small files' digest info
stream_load_pipe_count The total count of stream load pipes, each pipe has a limited buffer size (default 1M)
tablet_writer_count The total count of tablet writers
brpc_endpoint_stub_count The total count of brpc endpoints
```
1. MemTableFlushExecutor maintain a ThreadPool to receive FlushTask.
2. FlushToken is used to seperate different tasks from different tablets.
Every DeltaWriter of tablet constructs a FlushToken,
task in FlushToken are handle serially, task between FlushToken are
handle concurrently.
3. I have remove thread limit on data_dir, because of I/O is not the main
timer consumer of Flush thread. Much of time is consumed in CPU decoding
and compress.
Mainly contains the following modifications:
1. Use `std::unique_ptr` to replace some naked pointers
2. Modify some methods from member-method to local-static-function
3. Modify some methods do not need to be public to private
4. Some formatting changes: such as wrapping lines that are too long
5. Remove some useless variables
6. Add or modify some comments for easier understanding
No functional changes in this patch.
* [Alter Table] No need to check whether table is stable when doing some kinds of alter operation.
Not all alter table operation require table to be stable. Such as rename, modify meta data.
Mainly fix the following issues:
1. A null pointer exception is raised when a database or table is dropped. The expected behavior is that the routine load job is stopped.
2. Memory leaks. Batch routine load task submissions are no longer performed, and modifications are submitted separately for each task.
3. Unreasonable task timeout.
Routine load tasks should not be queued in the BE thread pool for execution. The task sent to the BE should be executed immediately, otherwise the task in the FE will be timeout first. Eventually leads to constant timeout for all subsequent tasks.
4. All routine load job should be scheduled once it being submitted. Not waiting the available BE slot. Otherwise, all later submitted jobs may not be scheduled forever.
Remove the default constructor for UniqueID
Add a gen_uid method in UniqueId. If need to generate a new uid, users should call this api explicitly.
Reuse boost random generator not generate a new one every time.
Commit kafka offset in routine load
Kafka will decide whether to delete data based on whether all consumer group is commit offset or not. If there is no commit offset, the kafka server disk may be full
1. get_json_xxx() now support using quoto to escape dot
2. Implement json_path_prepare() function to preprocess json_path
Performance of get_json_string() on 1000000 rows reduces from 2.27s to 0.27s
1. Use a data consumer group to share a single stream load pipe with multi data consumers. This will increase the consuming speed of Kafka messages, as well as reducing the task number of routine
load job.
Test results：
* 1 consumer, 1 partitions:
consume time: 4.469s, rows: 990140, bytes: 128737139. 221557 rows/s, 28M/s
* 1 consumer, 3 partitions:
consume time: 12.765s, rows: 2000143, bytes: 258631271. 156689 rows/s, 20M/s
blocking get time(us): 12268241, blocking put time(us): 1886431
* 3 consumers, 3 partitions:
consume time(all 3): 6.095s, rows: 2000503, bytes: 258631576. 328220 rows/s, 42M/s
blocking get time(us): 1041639, blocking put time(us): 10356581
The next 2 cases show that we can achieve higher speed by adding more consumers. But the bottle neck transfers from Kafka consumer to Doris ingestion, so 3 consumers in a group is enough.
I also add a Backend config `max_consumer_num_per_group` to change the number of consumers in a data consumer group, and default value is 3.
In my test(1 Backend, 2 tablets, 1 replicas), 1 routine load task can achieve 10M/s, which is same as raw stream load.
2. Add OFFSET_BEGINNING and OFFSET_END support for Kafka routine load
