Show placement information in SHOW TABLE x REGIONS

[morgo]

Placement rules are by design asyncronously scheduled. The state of scheduling can be observed in the statement SHOW PLACEMENT FOR.

This is a feature request to also include current placement information in the statement SHOW TABLE x REGIONS. The motivation for doing this is as follows:

• It allows observability of the current scheduling state (show placement for is too binary). Consider this as similar to how we can observe the progress of DDL
• It allows for verification that the rules are in fact followed correctly. We can incorporate this into our own scenario testing, but also for users that have compliance or HA requirements they might script against it to verify correctness.

The current output of SHOW TABLE x REGIONS is as follows:

tidb> show table t1 regions;
+-----------+---------------+---------------+-----------+-----------------+-------+------------+---------------+------------+----------------------+------------------+
| REGION_ID | START_KEY     | END_KEY       | LEADER_ID | LEADER_STORE_ID | PEERS | SCATTERING | WRITTEN_BYTES | READ_BYTES | APPROXIMATE_SIZE(MB) | APPROXIMATE_KEYS |
+-----------+---------------+---------------+-----------+-----------------+-------+------------+---------------+------------+----------------------+------------------+
|         4 |               | t_63_r_89565  |         5 |               1 | 5     |          0 |          4337 |      80007 |                  107 |           367623 |
|         6 | t_63_r_89565  | t_63_r_179845 |         7 |               1 | 7     |          0 |          1323 |   12187800 |                   91 |                0 |
|         8 | t_63_r_179845 | t_63_r_270126 |         9 |               1 | 9     |          0 |          1323 |   12187935 |                  102 |           115360 |
|        10 | t_63_r_270126 | t_63_r_360407 |        11 |               1 | 11    |          0 |            27 |   12187935 |                   91 |            63475 |
|        12 | t_63_r_360407 | t_63_r_450687 |        13 |               1 | 13    |          0 |          1323 |   12187800 |                  122 |           566573 |
|        14 | t_63_r_450687 | t_63_r_540968 |        15 |               1 | 15    |          0 |          1323 |    8125290 |                   95 |            57600 |
|        16 | t_63_r_540968 | t_63_r_631248 |        17 |               1 | 17    |          0 |          1323 |          0 |                  109 |           207822 |
|         2 | t_63_r_631248 |               |         3 |               1 | 3     |          0 |          1498 |          0 |                  103 |           104284 |
+-----------+---------------+---------------+-----------+-----------------+-------+------------+---------------+------------+----------------------+------------------+
8 rows in set (0.01 sec)

I propose that we add 2 columns:

• Scheduling Constraints: a list of recorded constraints for the region. It can be json formatted.
• Scheduling State: the same output as from SHOW PLACEMENT FOR.

[johnhaxx7]

Hi @morgo, I'm pretty interested in this task, but I have no idea of the complexity and effort, or any dependency for this.
If it's ok for me to take this, any chance you can give a brief walkthrough so that I can make sure things are on track?
I'm at PST timezone, a short chat on Slack would be great!

[morgo]

Hi @morgo, I'm pretty interested in this task, but I have no idea of the complexity and effort, or any dependency for this. If it's ok for me to take this, any chance you can give a brief walkthrough so that I can make sure things are on track? I'm at PST timezone, a short chat on Slack would be great!

Sure, you can find me on TiDB Community Slack :-)

The code for show placement is in ./executor/show_placement.go (copy this for how to get placement info). show table regions is in ./executor/show.go. The columns are defined here:

tidb/planner/core/planbuilder.go

Lines 2771 to 2784 in 173dd00

	func buildTableRegionsSchema() (*expression.Schema, types.NameSlice) {
	schema := newColumnsWithNames(11)
	schema.Append(buildColumnWithName("", "REGION_ID", mysql.TypeLonglong, 4))
	schema.Append(buildColumnWithName("", "START_KEY", mysql.TypeVarchar, 64))
	schema.Append(buildColumnWithName("", "END_KEY", mysql.TypeVarchar, 64))
	schema.Append(buildColumnWithName("", "LEADER_ID", mysql.TypeLonglong, 4))
	schema.Append(buildColumnWithName("", "LEADER_STORE_ID", mysql.TypeLonglong, 4))
	schema.Append(buildColumnWithName("", "PEERS", mysql.TypeVarchar, 64))
	schema.Append(buildColumnWithName("", "SCATTERING", mysql.TypeTiny, 1))
	schema.Append(buildColumnWithName("", "WRITTEN_BYTES", mysql.TypeLonglong, 4))
	schema.Append(buildColumnWithName("", "READ_BYTES", mysql.TypeLonglong, 4))
	schema.Append(buildColumnWithName("", "APPROXIMATE_SIZE(MB)", mysql.TypeLonglong, 4))
	schema.Append(buildColumnWithName("", "APPROXIMATE_KEYS", mysql.TypeLonglong, 4))
	return schema.col2Schema(), schema.names

Try getting started, and if you get stuck I can check out your branch and take a look.

[johnhaxx7]

Thank you Morgan! Assign per offline discussion

[johnhaxx7]

/assign