Modify the implementation of MemTracker:
1. Simplify a lot of useless logic;
2. Added MemTrackerTaskPool, as the ancestor of all query and import trackers, This is used to track the local memory usage of all tasks executing;
3. Add cosume/release cache, trigger a cosume/release when the memory accumulation exceeds the parameter mem_tracker_consume_min_size_bytes;
4. Add a new memory leak detection mode (Experimental feature), throw an exception when the remaining statistical value is greater than the specified range when the MemTracker is destructed, and print the accurate statistical value in HTTP, the parameter memory_leak_detection
5. Added Virtual MemTracker, cosume/release will not sync to parent. It will be used when introducing TCMalloc Hook to record memory later, to record the specified memory independently;
6. Modify the GC logic, register the buffer cached in DiskIoMgr as a GC function, and add other GC functions later;
7. Change the global root node from Root MemTracker to Process MemTracker, and remove Process MemTracker in exec_env;
8. Modify the macro that detects whether the memory has reached the upper limit, modify the parameters and default behavior of creating MemTracker, modify the error message format in mem_limit_exceeded, extend and apply transfer_to, remove Metric in MemTracker, etc.;
Modify where MemTracker is used:
1. MemPool adds a constructor to create a temporary tracker to avoid a lot of redundant code;
2. Added trackers for global objects such as ChunkAllocator and StorageEngine;
3. Added more fine-grained trackers such as ExprContext;
4. RuntimeState removes FragmentMemTracker, that is, PlanFragmentExecutor mem_tracker, which was previously used for independent statistical scan process memory, and replaces it with _scanner_mem_tracker in OlapScanNode;
5. MemTracker is no longer recorded in ReservationTracker, and ReservationTracker will be removed later;
For the first, we need to make a parameter to discribe the data is local or remote.
At then, we need to support some basic function to support the operation for remote storage.
The broker scan node has two tuple descriptors:
One is dest tuple and the other is src tuple.
The src tuple is used to read the lines of the original file,
and the dest tuple is used to save the converted lines.
The preceding filter is executed on the src tuple, so src tuple descriptor should be used
to initialize the filter expression
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
1. insert very large string value may coredump
2. some analitic functiuon and agg function result may be incorrect
3. string compare may be coredump when string type is too large
4. string type in delete condition can not process correctly
5. add text/blob as alias of string to compitable with mysql
6. fix string type min/max agg may process incorrectly
fix#6269
The outline of our changes is to improve our memory in case of OOM in BE and to speed up the calculation.
1. We do not need to do Aggregation in load, which has already been done in the ETL spark job.
2. Based on 1, we do not need to serialize/deserialize bitmap/HLL objects.
The problem I want to solve is described in #6355.
This CL mainly changes:
1. Support compacting tablets under alter operations
On BE side, the compaction logic will select tablets which state is "TABLET_NOTREADY" to do cumulative compaction.
2. Remove "alter_task" field in tablet's meta on BE side.
"alter_task" field is never used long time ago
3. Support doing delete operation when table is doing alter operation.
Previously, when a table is doing alter operation, execution of delete will return error: Table's state is not NORMAL.
But now, delete can be executed successfully only if the condition column is not under schema change.
And delete condition will be applied to all materialized indexes.
* [Enhance] Make MemTracker more accurate (#5515)
This PR main about:
1. Improve the readability of MemTrackers' name
2. Add the MemTracker of:
* Load
* Compaction
* SchemaChange
* StoragePageCache
* TabletManager
3. Change SchemaChange to a Singleon
* revise some code for Code Review
* change the name of mem_tracker
* keep reader_context have the same lifetime of rowset_reader in schema change.
* change vlog notice to log(warning) in schema change
Support conditional filtering of original data in broker load and routine load
eg:
```
LOAD LABEL `label1`
(
DATA INFILE ('bos://cmy-repo/1.csv')
INTO TABLE tbl2
COLUMNS TERMINATED BY '\t'
(event_day, product_id, ocpc_stage, user_id)
SET (
ocpc_stage = ocpc_stage + 100
)
PRECEDING FILTER user_id = 1381035
WHERE ocpc_stage > 30
)
...
```
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
This CL mainly changes:
1. Avoid repeated sending of common components in Fragments
In the previous implementation, a query may generate multiple Fragments,
these Fragments contain some common information, such as DescriptorTable.
Fragment will be sent to BE in a certain order, so these public information will be sent repeatedly
and generated repeatedly on the BE side.
In some complex SQL, these public information may be very large,
thereby increasing the execution time of Fragment.
So I improved this. For multiple Fragments sent to the same BE, only the first Fragment will carry
these public information, and it will be cached on the BE side, and subsequent Fragments
no longer need to carry this information.
In the local test, the execution time of some complex SQL can be reduced from 3 seconds to 1 second.
2. Add the time-consuming part of FE logic in Profile
Including SQL analysis, planning, Fragment scheduling and sending on the FE side, and the time to fetch data.
For[Spark Load]
1 support decimal andl largeint
2 add validate logic for char/varchar/decimal
3 check data load from hive with strict mode
4 support decimal/date/datetime aggregator
Since the Segment V2 has been released for a long time, we should make it as default storage format for newly created table.
This CL mainly changes:
1. For all newly created tables, their default storage format is Segment V2.
2. For all already exist tablets, their storage format remain unchanged.
3. Fix bugs described in Fix#4384 and Fix#4385
After PR: #4135, If a mem tracker has parent, it should be created by 'CreateTracker'.
So I removed other unused constructors.
And also fix the bug described in #4344
1、Add a PushBrokerReader in push_handle.cpp.
2、PushBrokerReader wraps the ParquetScanner to support reading data from parquet format file through broker.
Support compaction operation to compact only one rowset.
After the modification, the last rowset of the tablet will
also be compacted.
At the same time, we added a `segments_overlap_pb` field to
the rowset meta. Used to describe whether the segment data
in the rowset overlaps. This field is set by `rowset_writer`.
Initially UNKNOWN for compatibility with existing data.
In addition, the version hash of the rowset generated after
compaction is directly set to the version hash of last rowset
participating in compaction, to ensure that the tablet's
version hash remains unchanged after compaction.
BetaRowsetWriter is used to write rowset in V2 segment format.
This PR contains several interface changes
1. Rowset.make_snapshot() is renamed to `link_files_to` because hard links are also useful in copy task, linked schema change, etc
2. Rowset.copy_files_to_path() is renamed to `copy_files_to` to be consistent with other names
3. RowsetWriter.mem_pool() is removed because not all rowset writers use MemPool
4. RowsetWriter.garbage_collection() is removed because it's not used by clients
5. SegmentGroup's make_snapshot() is removed because link_segments_to_path() provides similar functionality
Currently, we have Field and ColumnSchema to access column data in a
row. These two classes are mostly the same. So we should unify these to
one class. Now, Field has offset information, which is an row attribute,
so we remove offset in Field.
RowCursor now has some logic which belong to Schema, so in this patch I
add Schema attribute to RowCursor to make RowCursor simple. After this
change, only Schema will handle Field/ColumnSchema.
I extract some logic from RowCursor to be/src/olap/row.h, then we can
use same logic to handle different types of row. Each type of row has
same function that to get Cell of this row. A cell represent a column
content with a null indicator.
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.
1. No one can set root password expect for root user itself
2. NODE_PRIV cannot be granted.
3. ADMIN_PRIV and GRANT_PRIV can only be granted or revoked on *.*
4. No one can modifly privs of default role 'operator' and 'admin'.
5. No user can be granted to role 'operator'.
Fixed: the running load limit should not be applied to replay logic. It will cause replay or loading image fail.
Changed: optimize the problem of too many directories under mini load directory.
Fixed: missing password and auth check when handling mini load request in Frontend.
Fixed: DomainResolver should start after Frontends transfer to a certain ROLE, not in Catalog construction methods.
Fixed: a stupid bug that no one can set password for root user... fix it: only root user can set password for root.
Fixed: read null data twice
When reading data with a null value, in some cases, the same data will be read twice by the storage engine,
resulting in a wrong result.The reason for this problem is that when splitting,
and the start key is the minimum value, the data with null is read.
Fixed: add a flag to prevent DomainResovler thread start twice.
Fixed: fixed a mem leak of using ByteBuf when parsing auth info of http request.
Fixed: add a new config 'disable_hadoop_load', default is false, set to true to disable hadoop load.
Changed: add detail error msg of submitting hadoop load job in show load result.
Fixed: Backend process should be crashed if failed to saving header.
Added: exposure backend info to user when encounter error on Backend. for debugging it more convenient.
Fixed: Should remove fd from map when inputstream or outputstream is closed in Broker process.
Fixed: Change all files' LF to unix format.
Internal commit id: merge from dfcd0aca18eed9ff99d188eb3d01c60d419be1b8
