In some cloud native deployment scenario, BE(especially the Compute Node BE) will be add to cluster and remove from cluster very frequently. User's query will fail if there is a fragment is running on the shutting down BE. Users could use stop_be.sh --grace, then BE will wait all running queries to stop to avoiding running query failure, but if the waiting time exceed the limit, then be will exit directly. During this period, FE will not send any queries to BE and waiting for all running queries to stop
Current initialization dependency:
Daemon ───┬──► StorageEngine ──► ExecEnv ──► Disk/Mem/CpuInfo
│
│
BackendService ─┘
However, original code incorrectly initialize Daemon before StorageEngine.
This PR also stop and join threads of daemon services in their dtor, to ensure Daemon services release resources in reverse order of initialization via RAII.
now,`hostname_to_ip` only can resolve `ipv4`,Therefore, a method is provided to parse ipv4 or ipv6 based on parameters。
when `_heartbeat` call `hostname_to_ip`,Resolve to ipv4 or ipv6, determined by `BackendOptions.is_bind_ipv6` Decision
Additionally, a method is provided to first attempt to parse the host into ipv4, and then try ipv6 if it fails
When FE is deployed on a virtual machine and CN is deployed on k8s, FE needs to use a proxy IP to communicate with
CN nodes, and BE cannot resolve the proxy IP from the local network card.
We change the previous verification rules and obey the IP assigned by the master
Main changes:
1. If fqdn is enabled in the configuration file, when fe starts, localAddr will obtain fqdn instead of IP, priority_ Networks will fail
2. The IP and host names of Backend and Front are combined into one field, host. When fqdn is enabled, it represents the host name, and when not enabled, it represents the IP address
3. The communication between clusters directly uses fqdn, and various Connection pool add authentication mechanisms to prevent the IP address of the domain name from changing and the connection between nodes from making errors
4. No longer requires polling to verify if the IP has changed, delete fqdnManager
5. Change the method of verifying the legitimacy of nodes between FEs from obtaining client IP to displaying the identity of the transmitting node itself in the HTTP request header or the message body of the throttle
6. When processing the heartbeat, if BE finds that the host stored by itself is inconsistent with the host stored by the master, after verifying the legitimacy of the host, it will change its own host instead of directly reporting an error
7. Simplify the generation logic of fe name
Scope of influence:
1. Establishing communication connections between clusters
2. Determine whether it is the same node through attributes such as IP
3. Print Log
4. Information display
5. Address Splicing
6. k8s deployment
7. Upgrade compatibility
Test plan:
1. Change the IP address of the node, while keeping the fqdn unchanged, change the IP addresses of fe and be, and verify whether the cluster can read and write data normally
2. Use the master code to generate metadata, and use the previous metadata on the current pr to verify whether it is compatible with the old version (upgrading is no longer supported if fqdn has been enabled before)
3. Deploy fe and be clusters using k8s to verify whether the cluster can read and write data normally
4. According to https://doris.apache.org/zh-CN/docs/dev/admin-manual/cluster-management/fqdn?_highlight=fqdn#%E6%97%A7%E9%9B%86%E7%BE%A4%E5%90%AF%E7%94%A8fqdn Upgrading old clusters
5. Use streamload to specify the fqdn of fe and be to import data separately
6. Use different users to start transactions and write data using insert statements
Currently, there are some useless includes in the codebase. We can use a tool named include-what-you-use to optimize these includes. By using a strict include-what-you-use policy, we can get lots of benefits from it.
Fix some logic about broker load using new file scanner, with parquet format:
1. If columns are specified in load stmt, but none of them are in parquet file,
error will be thrown like `err: No columns found in file`. See `parquet_s3_case4`
2. If the first column of table are not in table, the result number of rows is wrong.
See `parquet_s3_case8`
3. If column specified in `columns` in load stmt does not exist in file and table,
error will be thrown like: `failed to find default value expr for slot: x1`. See `parquet_s3_case2`
There are 3 error code types in BE: OLAPStatus AgentStatus Status.
It is very confused and sometimes conflict during write code.
I will try to unify them to Status.
1. replace all boost::shared_ptr to std::shared_ptr
2. replace all boost::scopted_ptr to std::unique_ptr
3. replace all boost::scoped_array to std::unique<T[]>
4. replace all boost:thread to std::thread
We use 'LastStartTime' in backends list to check whether there is an unexpected
restart of BE, but it will be changed as BE's first heartbeat time after FE
restarted, it would be better to set it to BE's actual start time.
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.
Fix#3390
This CL add more info in `JobDetails` column of `SHOW LOAD` result for Broker Load Job.
For example:
```
{
"Unfinished backends": {
"9c3441027ff948a0-8287923329a2b6a7": [10002]
},
"All backends": {
"9c3441027ff948a0-8287923329a2b6a7": [10002, 10004, 10006]
},
"ScannedRows": 2390016,
"TaskNumber": 1,
"FileNumber": 1,
"FileSize": 1073741824
}
```
2 newly added keys:
`Unfinished backends` indicates the BE which task on them are not finished.
`All backends` indicates the BE which this job has tasks on it.
One more thing, I pass the Backend Id along with the heartbeat msg from FE to BE, so that BE can
know the Id of themselves.
In a large scale cluster, we may rolling upgrade BEs, this patch add a
column named 'Version' for command 'show backends;', as well as website
'/system?path=//backends', to provide a method to check whether there
is any BE missing upgraded.
Currently, the report from BE to FE is completed in the background
threads of `AgentServer` (`report_tablet_thread` and
`report_disk_stat_thread`). These two threads will sleep and be in
a standby state after each report, if there is any need to report
immediately, they will be notified and wake up immediately to report.
For example, when background thread (`disk_monitor_thread`) in
`StorageEngine` finds some tablets were deleted, it will notify
`AgentServer` to trigger a report immediately.
In the current implementation, in order to report ASAP, a local variable
(`_is_drop_tables`) and two other flags are used to record whether
reporting is needed, and then `StorageEngine::disk_monitor_thread` checks
the value of this variable every time it runs, to determine whether it
needs to be triggered Reporting. This is actually superfluous, and it
may result in untimely notifications, as shown below:
```
(thread_1) (thread_2)
disk-monitor disk-stat-reporter
| |
| reporting
| |
notify_1 |
| |
| wait_for_notify(will wait until timeout or next notification)
| |
V V
```
When `report_tablet_thread` has not started waiting,
`StorageEngine::disk_monitor_thread` triggers a notification, so this
notification will not be received by `report_tablet_thread`,
resulting in the BE not reporting to the FE until the lock times out
or the next round of `disk_monitor_thread` detection.
This change restructures the triggering implementation, and solves the above problem.
This change also changes some methods(that do not need to be public) to private.
The control framework is implemented through heartbeat message. Use uint64_t as flags to control different functions.
Now add a flag to set the default rowset type to beta.
NOTE: This patch would modify all Backend's data.
And this will cause a very long time to restart be.
So if you want to interferer your product environment,
you should upgrade backend one by one.
1. Refactoring be is to clarify the structure the codes.
2. Use unique id to indicate a rowset.
Nameing rowset with tablet_id and version will lead to
many conflicts among compaction, clone, restore.
3. Extract an rowset interface to encapsulate rowsets
with different format.
Add path info of replica in catalog
Also fix a bug that when calling check_none_row_oriented_table,
store is null, it cannot be used to create table.
Instead, OLAPHeader can be used to get storage type information.
* Reduce UT binary size
Almost every module depend on ExecEnv, and ExecEnv contains all
singleton, which make UT binary contains all object files.
This patch seperate ExecEnv's initial and destory to anthor file to
avoid other file's dependence. And status.cc include debug_util.h which
depend tuple.h tuple_row.h, and I move get_stack_trace() to
stack_util.cpp to reduce status.cc's dependence.
I add USE_RTTI=1 to build rocksdb to avoid linking librocksdb.a
Issue: #292
* Update
Added: Support getting column size and precision info of table or view using JDBC.
Updated: Change the promethues type name GAUGE to lowercase, to fit the latest promethues version.
Updated: Backend ip saved in FE will be compared with BE's local ip when doing heartbeat, to avoid false positive heartbeat response.
Updated: Using version_num of tablet instead of calculating nice value to select cumulative compaction candicates.
Fixed: Predicates should not be pushed down to subquery which contains limit clause.
Fixed: Fix the formula of calculating BE load score.
Fixed: Fix a bug that in some edge cases, non-master Fontend may wait for a unnecessary long timeout after forwarding cmd to Master FE.
Fixed: A bug that granting privs on more than one table does not work.
Fixed: Support 'Insert into' table which contains HLL columns.
Fixed: ExportStmt' toSql() method may throw NullPointer Exception if table does not exist.
Fixed: Remove unnecessary 'get capacity' operation to avoid IO impact.
Internal commit id: merge to c16bd603a53dfe2089ff95704c698a738c317792
1. Apache HDFS broker support HDFS HA and Hadoop kerberos authentication.
2. New Backup and Restore function. Use Fs Broker to backup your data to HDFS or restore them from HDFS.
3. Table-Level Privileges. Grant fine-grained privileges on table-level to specified user.
4. A lot of bugs fixed.
5. Performance improvement.
1. Implement Backend http server using libevent instead of mongoose.
2. Remove Old Hypertable rpc framework, use brpc instead.
3. Change rpc from FE to BE to brpc.
4. Fs broker support HDFS HA.
5. add more metrics to monitor.
6. Lots of bug fixed.
