use this stmt to show tablets storage format in be, if verbose is set,
will show detail message of tablet storage format.
e.g.
```
MySQL [(none)]> admin show tablet storage format;
+-----------+---------+---------+
| BackendId | V1Count | V2Count |
+-----------+---------+---------+
| 10002 | 0 | 2867 |
+-----------+---------+---------+
1 row in set (0.003 sec)
MySQL [test_query_qa]> admin show tablet storage format verbose;
+-----------+----------+---------------+
| BackendId | TabletId | StorageFormat |
+-----------+----------+---------------+
| 10002 | 39227 | V2 |
| 10002 | 39221 | V2 |
| 10002 | 39215 | V2 |
| 10002 | 39199 | V2 |
+-----------+----------+---------------+
4 rows in set (0.034 sec)
```
add storage format infomation to show full table statment.
```
MySQL [test_query_qa]> show full tables;
+-------------------------+------------+---------------+
| Tables_in_test_query_qa | Table_type | StorageFormat |
+-------------------------+------------+---------------+
| bigtable | BASE TABLE | V2 |
| test_dup | BASE TABLE | V2 |
| test | BASE TABLE | V2 |
| baseall | BASE TABLE | V2 |
| test_string | BASE TABLE | V2 |
+-------------------------+------------+---------------+
5 rows in set (0.002 sec)
```
Buffer flip is used incorrectly.
When the hash key is string type, the hash value is always zero.
The reason is that the buffer of string type is obtained by wrap, which is not needed to flip.
If we do so, the buffer limit for read will be zero.
* load newly generated image file as soon as generated to check if it is valid.
* delete the latest invalid image file
* fix
* fix
* get filePath from saveImage() to ensure deleting the correct file while exception happens
* fix
Co-authored-by: wuhangze <wuhangze@jd.com>
* avoiding a corrupt image file when there is image.ckpt with non-zero size
For now, saveImage writes data to image.ckpt via an append FileOutputStream,
when there is a non-zero size file named image.ckpt, a disaster would happen
due to a corrupt image file. Even worse, fe only keeps the lastest image file
and removes others.
BTW, image file should be synced to disk.
It is dangerous to only keep the latest image file, because an image file is
validated when generating the next image file. Then we keep an non validated
image file but remove validated ones. So I will issue a pr which keeps at least
2 image file.
* append other data after MetaHeader
* use channel.force instead of sync
For now, dbTransactionManager::getTransactionNum is only used by
checkpoint to get transaction num to put into a image file. However,
transactions written into a image file do not come from the same
data structure as the num comes. Thus, we should pay much attention to
assure two data structue is consistent on size. Actually, it is
very difficult to do so.
This patch just let getTransactionNum get number from the same data
structure as write method.
The change was introduced by b93e841688.
This is the first PR for statistics collection includes some implementations of the statistics(#6370), it will not affect any existing code and users will not be able to create statistics job.
It mainly implements the semantic checking module for statistical information collection jobs, and the job creation module.
The syntax is:
ANALYZE [[ db_name.tb_name ] [( column_name [, ...] )], ...] [ PROPERTIES(...) ]
e.g.
ANALYZE;
ANALYZE tbl1;
ANALYZE tbl1(col1, col2) PROPERTIES("cbo_ statistics_ task_ timeout" = "10");
Two configurations have been added:
Timeout time of a single task max_cbo_statistics_task_timeout_sec
The maximum number of running jobs the system can receive cbo_max_statistics_job_num
Co-authored-by: weizhengte <1141550741@qq.com>
Co-authored-by: weizhengte <weizhengte@foxmail.com>
Co-authored-by: EmmyMiao87 <522274284@qq.com>
Co-authored-by: frankywei <frankywei@tencent.com>
- Add two new types to stream load boker load: **csv_with_names** and **csv_with_name_sand_types**
- Add two new types to export: **csv_with_names** and **csv_with_names_and_types**
* [Forbidden](Vec) Switch to non-vec engine when outer join + not null column
Vectorized code will occur `core` in the case of ```outer join + not null column```, such as issue #7901
So we need to fall back from vectorized mode to non-vectorized mode when we encounter this situation.
If the nullside column of the outer join is a column that must return non-null like count(*)
then there is no way to force the column to be nullable.
At this time, vectorization cannot support this situation,
so it is necessary to fall back to non-vectorization for processing.
For example:
Query: set enable_vectorized_engine=true
Query: select * from t1 left join (select k1, count(k2) as count_k2 from t2 group by k1) tmp on t1.k1=tmp.k1
Result: Query goes non-vectorized engine
