Choose synchronous nodes based on replication lag (#1786)

This commit makes it possible to configure the maximum lag (`maximum_lag_on_syncnode`) after which Patroni will "demote" the node from synchronous and replace it with another node.

The previous implementation always tried to stick to the same synchronous nodes (even if they are not optimal ones).
This commit is contained in:
krishna
2021-02-02 15:45:02 +01:00
committed by GitHub
parent 9d7d4423e3
commit b3dc765e6d
9 changed files with 92 additions and 23 deletions
+19
View File
@@ -28,6 +28,25 @@ Feature: basic replication
When I issue a GET request to http://127.0.0.1:8009/async
Then I receive a response code 200
Scenario: check stuck sync replica
Given I issue a PATCH request to http://127.0.0.1:8008/config with {"maximum_lag_on_syncnode": 15000000, "postgresql": {"parameters": {"synchronous_commit": "remote_apply"}}}
Then I receive a response code 200
And I create table on postgres0
And table mytest is present on postgres1 after 2 seconds
And table mytest is present on postgres2 after 2 seconds
When I pause wal replay on postgres2
And I load data on postgres0
Then "sync" key in DCS has sync_standby=postgres1 after 15 seconds
And I resume wal replay on postgres2
And I sleep for 2 seconds
And I issue a GET request to http://127.0.0.1:8009/sync
Then I receive a response code 200
When I issue a GET request to http://127.0.0.1:8010/async
Then I receive a response code 200
When I issue a PATCH request to http://127.0.0.1:8008/config with {"maximum_lag_on_syncnode": -1, "postgresql": {"parameters": {"synchronous_commit": "on"}}}
Then I receive a response code 200
And I drop table on postgres0
Scenario: check multi sync replication
Given I issue a PATCH request to http://127.0.0.1:8008/config with {"synchronous_node_count": 2}
Then I receive a response code 200