*: implement inspection_summary system table which organizes metrics by link/module

[lonng]

What problem does this PR solve?

There are many metrics in the TiDB cluster and the users have no idea to use which metrics to locate the problem in the cluster because our metrics have not been organized in an efficient approach.

What is changed and how it works?

This PR try to introduce a system table inspection_summary, which is used to organize our metrics by read/write link or modules, e.g:

• read-link is used to organize all metrics in the reading query execution path via the top-bottom form.
• write-link is used to organize all metrics in the writting query execution path via the top-bottom form.
• wait-events is used to organize all duration metrics
• ddl/txn/rocksdb/raftstore...

NOTE

• Some predicates will be pushed down to InspectionSummaryRetriever, which means there is no extra cost in unused rules, e.g: select * from inspection_summary where rule='read-link' only retrieve metrics related to the reading link. The supported push-down columns: rule, metric_name.
• The quantile can be specified as an arbitrary positive number(s) by the user. e.g: select * from inspection_summary where rule='read-link' and quantile=0.999 or select * from inspection_summary where rule='read-link' and quantile in (0.80, 0.90, 0.99, 0.999, 1).

TODO

The value columns in the result should be a Double(22,6), but there will be many changes to fix this issue. It will be addressed in the next PR because this PR is so huge and will lose reviewability as diff growth.

Check List

Tests

• Unit test
• Manual test (add detailed scripts or steps below)

mysql> select * from inspection_summary where rule='read-link' and quantile=0.99;
+-----------+-----------------+----------------------------------------+-----------------------------------------+----------+------------------------+------------------------+------------------------+
| RULE      | INSTANCE        | METRIC_NAME                            | LABEL                                   | QUANTILE | AVG_VALUE              | MIN_VALUE              | MAX_VALUE              |
+-----------+-----------------+----------------------------------------+-----------------------------------------+----------+------------------------+------------------------+------------------------+
| read-link | localhost:10080 | tidb_get_token_duration                |                                         |     0.99 |                      0 |                      0 |                      0 |
| read-link | localhost:10080 | tidb_parse_duration                    | general                                 |     0.99 |                      0 |                      0 |                      0 |
| read-link | localhost:10080 | tidb_parse_duration                    | internal                                |     0.99 | 0.00024318023875114783 | 0.00015546666666666671 | 0.00031222857142857124 |
| read-link | localhost:10080 | tidb_compile_duration                  | general                                 |     0.99 |  0.0005602036363636362 | 0.00031885714285714286 |  0.0008511999999999967 |
| read-link | localhost:10080 | tidb_compile_duration                  | internal                                |     0.99 |  0.0005602036363636362 | 0.00031885714285714286 |  0.0008511999999999967 |
| read-link |                 | pd_tso_rpc_duration                    |                                         |     0.99 |  0.0006046491087783595 |               0.000495 |  0.0016449999999999941 |
| read-link |                 | pd_tso_wait_duration                   |                                         |     0.99 |  0.0007610674231760213 |  0.0004950000000000001 |  0.0019399999999999817 |
| read-link | localhost:10080 | tidb_execute_duration                  | general                                 |     0.99 |                      0 |                      0 |                      0 |
| read-link | localhost:10080 | tidb_execute_duration                  | internal                                |     0.99 |   0.006195393939393939 |    0.00432000000000001 |   0.009813333333333346 |
...
| read-link | localhost:20180 | tikv_per_read_max_bytes                | bytes_per_read_percentile95, raft       |        0 |                      0 |                      0 |                      0 |
| read-link | localhost:20180 | tikv_per_read_max_bytes                | bytes_per_read_percentile99, kv         |        0 |      9.931350874431612 |                9.92375 |      9.998446601941748 |
| read-link | localhost:20180 | tikv_per_read_max_bytes                | bytes_per_read_percentile99, raft       |        0 |                      0 |                      0 |                      0 |
| read-link | localhost:20180 | tikv_per_read_max_bytes                | bytes_per_read_standard_deviation, kv   |        0 |      9.116433007264865 |       8.65275822534578 |     13.146375472013052 |
| read-link | localhost:20180 | tikv_per_read_max_bytes                | bytes_per_read_standard_deviation, raft |        0 |                      0 |                      0 |                      0 |
+-----------+-----------------+----------------------------------------+-----------------------------------------+----------+------------------------+------------------------+------------------------+
387 rows in set (0.21 sec)

Release note

• Add a new system table inspection_summary which is used to organizes metrics by link/module

[Deardrops]

LGTM

[crazycs520]

LGTM

[sre-bot]

/run-all-tests