infoschema: resolve loopback address of node (#20546)

[ti-srebot]

cherry-pick #20546 to release-4.0

---

Signed-off-by: crazycs520 [email protected]

What problem does this PR solve?

fix #20528

When Using tiup to deploy a TiDB cluster, the ip of TiKV status-address is 0.0.0.0. then query the cluster_config system will return error.

information_schema> select * from cluster_info;
+------+----------------------+-----------------------+--------------+------------------------------------------+---------------------------+--------------+
| TYPE | INSTANCE             | STATUS_ADDRESS        | VERSION      | GIT_HASH                                 | START_TIME                | UPTIME       |
+------+----------------------+-----------------------+--------------+------------------------------------------+---------------------------+--------------+
| tidb | 192.168.220.171:4000 | 192.168.220.171:10080 | 4.0.0-beta.2 | 79327861bad9b91cca3aa8972e6daaa34bcea526 | 2020-10-20T20:10:56+08:00 | 12.676844s   |
| tidb | 10.0.1.8:4000        | 10.0.1.8:10080        | 4.0.0-beta.2 | 937949f44d0d25e21f5a5722dbb334773de38db9 | 2020-10-20T20:09:33+08:00 | 1m35.676847s |
| pd   | 10.0.1.8:2379        | 10.0.1.8:2379         | 4.1.0-alpha  | 4252d7efe4b1f0f32421f2ffd142b540e6396944 | 2020-10-20T20:09:26+08:00 | 1m42.676849s |
| tikv | 10.0.1.8:20160       | 0.0.0.0:20180         | 4.1.0-alpha  | 482bd12170a1802941bbd1ceed37c0672341c4ab | 2020-10-20T20:09:28+08:00 | 1m40.67685s  |
+------+----------------------+-----------------------+--------------+------------------------------------------+---------------------------+--------------+
information_schema> select * from cluster_config where type='tikv' and `key` like '%pool%' limit 2;
+------+----------+-----+-------+
| TYPE | INSTANCE | KEY | VALUE |
+------+----------+-----+-------+
0 rows in set
Time: 0.027s
information_schema> show warnings;
+---------+------+----------------------------------------------------------------------------------------+
| Level   | Code | Message                                                                                |
+---------+------+----------------------------------------------------------------------------------------+
| Warning | 1105 | Get "http://0.0.0.0:20180/config": dial tcp 0.0.0.0:20180: connect: connection refused |
+---------+------+----------------------------------------------------------------------------------------+
1 row in set

What is changed and how it works?

Normally, start tikv with specified advertise-status-addr parameter will be ok, but advertise-status-addr doesn't exist before tikv/tikv#7944, so TiUP can't specify the advertise-status-addr by default.

So, I have to fix this in TiDB. In this PR:

information_schema> select * from cluster_info;
+------+----------------------+-----------------------+--------------+------------------------------------------+---------------------------+--------------+
| TYPE | INSTANCE             | STATUS_ADDRESS        | VERSION      | GIT_HASH                                 | START_TIME                | UPTIME       |
+------+----------------------+-----------------------+--------------+------------------------------------------+---------------------------+--------------+
| tidb | 192.168.220.171:4000 | 192.168.220.171:10080 | 4.0.0-beta.2 | 41e29e8f3351896ddef8f33e587c280cb68ac658 | 2020-10-20T20:11:27+08:00 | 17.601069s   |
| tidb | 10.0.1.8:4000        | 10.0.1.8:10080        | 4.0.0-beta.2 | 937949f44d0d25e21f5a5722dbb334773de38db9 | 2020-10-20T20:09:33+08:00 | 2m11.601073s |
| pd   | 10.0.1.8:2379        | 10.0.1.8:2379         | 4.1.0-alpha  | 4252d7efe4b1f0f32421f2ffd142b540e6396944 | 2020-10-20T20:09:26+08:00 | 2m18.601074s |
| tikv | 10.0.1.8:20160       | 10.0.1.8:20180        | 4.1.0-alpha  | 482bd12170a1802941bbd1ceed37c0672341c4ab | 2020-10-20T20:09:28+08:00 | 2m16.601075s |
+------+----------------------+-----------------------+--------------+------------------------------------------+---------------------------+--------------+
4 rows in set
Time: 0.018s
information_schema> select * from cluster_config where type='tikv' and `key` like '%pool%' limit 1;
+------+----------------+---------------------------+-------+
| TYPE | INSTANCE       | KEY                       | VALUE |
+------+----------------+---------------------------+-------+
| tikv | 10.0.1.8:20160 | raftstore.apply-pool-size | 2     |
+------+----------------+---------------------------+-------+
1 row in set
Time: 0.016s
information_schema> show warnings;
+-------+------+---------+
| Level | Code | Message |
+-------+------+---------+
0 rows in set

Related changes

• Need to cherry-pick to the release branch

Check List

Tests

• Unit test
• Manual test

Side effects

• No

Release note

• Fix issue of status address is a loopback address when querying system cluster tables.

[ti-srebot]

/run-all-tests

[wjhuang2016]

@crazycs520 Please resolve the conflict.

[crazycs520]

/run-all-tests

[crazycs520]

/run-sqllogic-test-2
/run-unit-test

[wjhuang2016]

LGTM

[lysu]

LGTM

[winoros]

/merge

[ti-srebot]

Sorry @winoros, this branch cannot be merged without an approval of release maintainers

[tiancaiamao]

/run-unit-test

[tiancaiamao]

/merge

[ti-srebot]

Sorry @tiancaiamao, this branch cannot be merged without an approval of release maintainers

[bb7133]

/merge

[ti-srebot]

Sorry @bb7133, this branch cannot be merged without an approval of release maintainers

[bb7133]

/merge

[ti-srebot]

/run-all-tests