infoschema: add metric database/table to query cluster metric table. #13757

crazycs520 · 2019-11-26T13:11:35Z

What problem does this PR solve?

Add add metric database/table to query cluster metric table.
Add 2 session variable to control the query metric time range:
- tidb_metric_schema_range_duration, the query duration, it will affect the generated promQL.
- tidb_metric_schema_step, the query step time, it will affect the step value when query promQL.

release test PR:https://github.com/pingcap/tidb-test/pull/968, merge test first.

mysql>use METRIC_SCHEMA;
mysql>select * from up;
+----------------------------+-------+-----------------+-------------------+
| time                       | value | instance        | job               |
+----------------------------+-------+-----------------+-------------------+
| 2019-12-04 18:46:33.398000 | 1.0   | 127.0.0.1:10080 | tidb              |
| 2019-12-04 18:46:33.398000 | 0.0   | 127.0.0.1:10081 | tidb              |
| 2019-12-04 18:46:33.398000 | 0.0   | 127.0.0.1:10082 | tidb              |
| 2019-12-04 18:46:33.398000 | 0.0   | 127.0.0.1:10083 | tidb              |
| 2019-12-04 18:46:33.398000 | 0.0   | 127.0.0.1:10090 | tidb              |
| 2019-12-04 18:46:33.398000 | 0.0   | 127.0.0.1:10091 | tidb              |
| 2019-12-04 18:46:33.398000 | 0.0   | 127.0.0.1:10092 | tidb              |
| 2019-12-04 18:46:33.398000 | 0.0   | 127.0.0.1:10093 | tidb              |
| 2019-12-04 18:46:33.398000 | 1.0   | 127.0.0.1:20180 | tikv              |
| 2019-12-04 18:46:33.398000 | 0.0   | 127.0.0.1:20181 | tikv              |
| 2019-12-04 18:46:33.398000 | 0.0   | 127.0.0.1:20182 | tikv              |
| 2019-12-04 18:46:33.398000 | 0.0   | 127.0.0.1:20183 | tikv              |
| 2019-12-04 18:46:33.398000 | 0.0   | 127.0.0.1:20184 | tikv              |
| 2019-12-04 18:46:33.398000 | 1.0   | 127.0.0.1:2379  | pd                |
| 2019-12-04 18:46:33.398000 | 0.0   | 127.0.0.1:2479  | pd                |
| 2019-12-04 18:46:33.398000 | 0.0   | 127.0.0.1:2579  | pd                |
| 2019-12-04 18:46:33.398000 | 0.0   | 127.0.0.1:9100  | overwritten-nodes |
| 2019-12-04 18:46:33.398000 | 1.0   | localhost:9090  | prometheus        |
+----------------------------+-------+-----------------+-------------------+
mysql>select * from query_duration;
+----------------------------+------------------+----------+----------+----------+
| time                       | value            | instance | sql_type | quantile |
+----------------------------+------------------+----------+----------+----------+
| 2019-12-04 18:46:37.134000 | 0.00430970543826 |          |          | 0.9      |
+----------------------------+------------------+----------+----------+----------+

You can use explain to check the generated promQL too:

mysql>desc select * from query_duration;
+----------------+----------+------+------------------------------------------------------------------------------------------------------------+
| id             | count    | task | operator info                                                                                              |
+----------------+----------+------+------------------------------------------------------------------------------------------------------------+
| MemTableScan_4 | 10000.00 | root | promQL:histogram_quantile(0.9, sum(rate(tidb_server_handle_query_duration_seconds_bucket{}[60s])) by (le)) |
+----------------+----------+------+------------------------------------------------------------------------------------------------------------+

What is changed and how it works?

Actually, we use promQL to query metric data. Every metric table has a related promQL.
Such as query_duration metric table, there is a definition in code:

	"query_duration": {
		promQL:   `histogram_quantile($QUANTILE, sum(rate(tidb_server_handle_query_duration_seconds_bucket{$LABEL_CONDITION}[$RANGE_DURATION])) by (le))`,
		labels:   []string{"instance", "sql_type"},
		quantile: 0.90,
	},
	"up": {
		promQL: `up{$LABEL_CONDITION}`,
		labels: []string{"instance", "job"},
	},

Then TiDB will generate a metric table as below:

mysql>desc query_duration;
+----------+-------------------+------+-----+-------------------+-------+
| Field    | Type              | Null | Key | Default           | Extra |
+----------+-------------------+------+-----+-------------------+-------+
| time     | datetime unsigned | YES  |     | CURRENT_TIMESTAMP |       |
| value    | double unsigned   | YES  |     | <null>            |       |
| instance | varchar(512)      | YES  |     | <null>            |       |
| sql_type | varchar(512)      | YES  |     | <null>            |       |
| quantile | double unsigned   | YES  |     | 0.9               |       |
+----------+-------------------+------+-----+-------------------+-------+
mysql>desc up;
+----------+-------------------+------+-----+-------------------+-------+
| Field    | Type              | Null | Key | Default           | Extra |
+----------+-------------------+------+-----+-------------------+-------+
| time     | datetime unsigned | YES  |     | CURRENT_TIMESTAMP |       |
| value    | double unsigned   | YES  |     | <null>            |       |
| instance | varchar(512)      | YES  |     | <null>            |       |
| job      | varchar(512)      | YES  |     | <null>            |       |
+----------+-------------------+------+-----+-------------------+-------+

As you can see, there is 3 variable in promQL, it will be replaced when execute SQL query:

$QUANTILE  
$LABEL_CONDITION
$RANGE_DURATION

If you execute below SQL:

select * from query_duration where time > "2019-11-25 00:00:00" and time < "2019-11-25 00:01:00" and sql_type="general";

It will generated a promQL as below:

histogram_quantile(0.90, sum(rate(tidb_server_handle_query_duration_seconds_bucket{sql_type="general"}[60s])) by (le))

The $QUANTILE and $RANGE_DURATION is 0.90, 60s, it's the session variable value of tidb_metric_schema_range_duration.

Then send the promQL to PD to query metric data which between in the "2019-11-25 00:00:00" and "2019-11-25 00:01:00".

Attension

Actually the variable in the promQL will always use the default value, use the value specified in SQL will be finished in the next PR.

Check List

Tests

Manual test (add detailed scripts or steps below)

Code changes

Side effects

Related changes

Release note

Write release note for bug-fix or new feature.

domain/domain.go

infoschema/infoschema_test.go

…table

codecov · 2019-12-04T10:03:40Z

Codecov Report

❗ No coverage uploaded for pull request base (master@eef8c39). Click here to learn what that means.
The diff coverage is 50.6172%.

@@             Coverage Diff             @@
##             master     #13757   +/-   ##
===========================================
  Coverage          ?   80.0891%           
===========================================
  Files             ?        487           
  Lines             ?     121431           
  Branches          ?          0           
===========================================
  Hits              ?      97253           
  Misses            ?      16435           
  Partials          ?       7743

infoschema/metricschema/table.go

sessionctx/variable/varsutil.go

…o metric-table

infoschema/metricschema/init.go

infoschema/metricschema/query.go

infoschema/metricschema/table.go

meta/autoid/autoid.go

sessionctx/variable/sysvar.go

djshow832

You missed IsMemOrSysDB in util/misc.go.

executor/cluster_reader.go

crazycs520 · 2019-12-19T08:29:31Z

/run-all-tests

crazycs520 · 2019-12-19T09:48:46Z

/run-all-tests tidb-test=pr/968

crazycs520 · 2019-12-19T10:21:12Z

/run-common-test tidb-test=pr/968

crazycs520 · 2019-12-19T10:29:25Z

/run-all-tests tidb-test=pr/968

crazycs520 · 2019-12-19T11:39:14Z

/run-integration-copr-test tidb-test=pr/968

crazycs520 · 2019-12-19T11:39:48Z

/run-common-test tidb-test=pr/968

…table

crazycs520 · 2019-12-19T14:33:20Z

/run-all-tests tidb-test=pr/968

djshow832

LGTM

…table

crazycs520 · 2019-12-20T08:01:39Z

/run-all-tests tidb-test=pr/968

…o metric-table

crazycs520 · 2019-12-20T08:03:07Z

/run-all-tests tidb-test=pr/968

crazycs520 · 2019-12-20T08:08:12Z

/run-integration-copr-test tidb-test=pr/968

pingcap#13757)

crazycs520 added 5 commits November 19, 2019 10:15

init

083a6b2

update query

c1abd6d

update query

33b943e

query metric from pd

899659d

remove debug print and query from pd

c4dcda6

crazycs520 added status/DNM type/usability labels Nov 26, 2019

lonng mentioned this pull request Nov 26, 2019

Tracking issue for TiDB built-in SQL Diagnostics #13567

Closed

56 tasks

crazycs520 added 3 commits November 27, 2019 14:18

Merge branch 'master' into metric-table

9faeeea

fix test

96a5ae7

fix lint

2ade77e

crazycs520 removed the status/DNM label Dec 4, 2019

lonng reviewed Dec 4, 2019

View reviewed changes

domain/domain.go Outdated Show resolved Hide resolved

lonng reviewed Dec 4, 2019

View reviewed changes

infoschema/infoschema_test.go Outdated Show resolved Hide resolved

crazycs520 added 2 commits December 4, 2019 16:57

address comment

58818e4

Merge branch 'master' of https://github.com/pingcap/tidb into metric-…

1a590aa

…table

crazycs520 added 2 commits December 9, 2019 11:11

Merge branch 'master' into metric-table

9f714ab

add explain info for metric table

ee7faa8

crazycs520 requested a review from a team as a code owner December 9, 2019 03:35

ghost requested review from alivxxx and lzmhhh123 and removed request for a team December 9, 2019 03:35

add explain info for metric table

81c8582

djshow832 reviewed Dec 9, 2019

View reviewed changes

infoschema/metricschema/table.go Outdated Show resolved Hide resolved

sessionctx/variable/varsutil.go Show resolved Hide resolved

crazycs520 added 2 commits December 9, 2019 13:05

address comment

4903109

Merge branch 'metric-table' of https://github.com/crazycs520/tidb int…

4b3f2e3

…o metric-table

djshow832 reviewed Dec 9, 2019

View reviewed changes

crazycs520 added 3 commits December 19, 2019 15:23

address comment

3a4d4ac

add fixme

e61271c

Merge branch 'master' into metric-table

67b6be0

lonng reviewed Dec 19, 2019

View reviewed changes

executor/cluster_reader.go Outdated Show resolved Hide resolved

address comment

35f71c4

Merge branch 'master' of https://github.com/pingcap/tidb into metric-…

2541b5a

…table

djshow832 reviewed Dec 20, 2019

View reviewed changes

djshow832 added the status/LGT1 Indicates that a PR has LGTM 1. label Dec 20, 2019

crazycs520 added 2 commits December 20, 2019 15:59

Merge branch 'master' of https://github.com/pingcap/tidb into metric-…

e93e254

…table

Merge branch 'master' into metric-table

ff8e2cf

crazycs520 added 2 commits December 20, 2019 16:02

fmt code

bb95818

Merge branch 'metric-table' of https://github.com/crazycs520/tidb int…

d393da6

…o metric-table

lonng approved these changes Dec 20, 2019

View reviewed changes

lonng merged commit 933715f into pingcap:master Dec 20, 2019

lonng added status/LGT2 Indicates that a PR has LGTM 2. and removed status/LGT1 Indicates that a PR has LGTM 1. labels Dec 20, 2019

XiaTianliang pushed a commit to XiaTianliang/tidb that referenced this pull request Dec 21, 2019

infoschema: add metric database/table to query cluster metric via SQL. (

924a1ec

pingcap#13757)

amyangfei mentioned this pull request Jan 6, 2020

Add one more TiDB system schema pingcap/tiflow#227

Merged

kennytm mentioned this pull request Feb 9, 2020

filter: use TiDB's IsMemOrSysDB() method in IsSystemSchema() pingcap/tidb-tools#316

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

infoschema: add metric database/table to query cluster metric table. #13757

infoschema: add metric database/table to query cluster metric table. #13757

crazycs520 commented Nov 26, 2019 •

edited

Loading

codecov bot commented Dec 4, 2019 •

edited

Loading

djshow832 left a comment

crazycs520 commented Dec 19, 2019

crazycs520 commented Dec 19, 2019

crazycs520 commented Dec 19, 2019

crazycs520 commented Dec 19, 2019

crazycs520 commented Dec 19, 2019

crazycs520 commented Dec 19, 2019

crazycs520 commented Dec 19, 2019

djshow832 left a comment

crazycs520 commented Dec 20, 2019

crazycs520 commented Dec 20, 2019

crazycs520 commented Dec 20, 2019

infoschema: add metric database/table to query cluster metric table. #13757

infoschema: add metric database/table to query cluster metric table. #13757

Conversation

crazycs520 commented Nov 26, 2019 • edited Loading

What problem does this PR solve?

What is changed and how it works?

Check List

codecov bot commented Dec 4, 2019 • edited Loading

Codecov Report

djshow832 left a comment

Choose a reason for hiding this comment

crazycs520 commented Dec 19, 2019

crazycs520 commented Dec 19, 2019

crazycs520 commented Dec 19, 2019

crazycs520 commented Dec 19, 2019

crazycs520 commented Dec 19, 2019

crazycs520 commented Dec 19, 2019

crazycs520 commented Dec 19, 2019

djshow832 left a comment

Choose a reason for hiding this comment

crazycs520 commented Dec 20, 2019

crazycs520 commented Dec 20, 2019

crazycs520 commented Dec 20, 2019

crazycs520 commented Nov 26, 2019 •

edited

Loading

codecov bot commented Dec 4, 2019 •

edited

Loading