gitops/patroni
1
0
Fork 0
mirror of https://github.com/outbackdingo/patroni.git synced 2026-08-25 14:53:37 +00:00
Code Issues Packages Projects Releases Wiki Activity

Don't retains member slots on nodes with nofailover tag (#3169)

Browse Source 
Followup on #3142
This commit is contained in:
Alexander Kukushkin
2024-09-17 11:21:54 +02:00
committed by GitHub
parent 87cb7481ae
commit d7e172c20a
3 changed files with 36 additions and 18 deletions
Download Patch File Download Diff File Expand all files Collapse all files
features/permanent_slots.feature
+6 -9
View File
@@ -7,12 +7,13 @@ Feature: permanent slots
Then I receive a response code 200
And Response on GET http://127.0.0.1:8008/config contains slots after 10 seconds
When I start postgres-1
And I start postgres-2
And I configure and start postgres-2 with a tag nofailover true
And I configure and start postgres-3 with a tag replicatefrom postgres-2
Then postgres-0 has a physical replication slot named test_physical after 10 seconds
And postgres-0 has a physical replication slot named postgres_1 after 10 seconds
And postgres-0 has a physical replication slot named postgres_2 after 10 seconds
And postgres-2 has a physical replication slot named postgres_3 after 10 seconds
And postgres-2 does not have a replication slot named test_physical
@slot-advance
Scenario: check that logical permanent slots are created
@@ -24,10 +25,9 @@ Feature: permanent slots
Scenario: check that permanent slots are created on replicas
Given postgres-1 has a logical replication slot named test_logical with the test_decoding plugin after 10 seconds
Then Logical slot test_logical is in sync between postgres-0 and postgres-1 after 10 seconds
And Logical slot test_logical is in sync between postgres-0 and postgres-2 after 10 seconds
And Logical slot test_logical is in sync between postgres-0 and postgres-3 after 10 seconds
And postgres-1 has a physical replication slot named test_physical after 2 seconds
And postgres-2 has a physical replication slot named test_physical after 2 seconds
And postgres-2 does not have a replication slot named test_logical
And postgres-3 has a physical replication slot named test_physical after 2 seconds
@slot-advance
@@ -36,9 +36,9 @@ Feature: permanent slots
And postgres-1 has a physical replication slot named postgres_0 after 2 seconds
And postgres-1 has a physical replication slot named postgres_2 after 2 seconds
And postgres-1 has a physical replication slot named postgres_3 after 2 seconds
And postgres-2 has a physical replication slot named postgres_0 after 2 seconds
And postgres-2 does not have a replication slot named postgres_0
And postgres-2 does not have a replication slot named postgres_1
And postgres-2 has a physical replication slot named postgres_3 after 2 seconds
And postgres-2 has a physical replication slot named postgres_1 after 2 seconds
And postgres-3 has a physical replication slot named postgres_0 after 2 seconds
And postgres-3 has a physical replication slot named postgres_1 after 2 seconds
And postgres-3 has a physical replication slot named postgres_2 after 2 seconds
@@ -50,11 +50,8 @@ Feature: permanent slots
And I get all changes from physical slot test_physical on postgres-0
Then Logical slot test_logical is in sync between postgres-0 and postgres-1 after 10 seconds
And Physical slot test_physical is in sync between postgres-0 and postgres-1 after 10 seconds
And Logical slot test_logical is in sync between postgres-0 and postgres-2 after 10 seconds
And Physical slot test_physical is in sync between postgres-0 and postgres-2 after 10 seconds
And Logical slot test_logical is in sync between postgres-0 and postgres-3 after 10 seconds
And Physical slot test_physical is in sync between postgres-0 and postgres-3 after 10 seconds
And Physical slot postgres_1 is in sync between postgres-0 and postgres-2 after 10 seconds
And Physical slot postgres_1 is in sync between postgres-0 and postgres-3 after 10 seconds
And Physical slot postgres_3 is in sync between postgres-2 and postgres-0 after 20 seconds
And Physical slot postgres_3 is in sync between postgres-2 and postgres-1 after 10 seconds
@@ -69,7 +66,7 @@ Feature: permanent slots
@slot-advance
Scenario: check that only non-permanent member slots are written to the retain_slots in /status key
And "status" key in DCS has postgres_0 in retain_slots
Given "status" key in DCS has postgres_0 in retain_slots
And "status" key in DCS has postgres_1 in retain_slots
And "status" key in DCS has postgres_2 in retain_slots
And "status" key in DCS does not have postgres_3 in retain_slots
patroni/dcs/__init__.py
+16 -9
View File
@@ -1116,16 +1116,23 @@ class Cluster(NamedTuple('Cluster',
def _get_members_slots(self, name: str, role: str, nofailover: bool,
can_advance_slots: bool) -> Dict[str, Dict[str, Any]]:
"""Get physical replication slots configuration for members that sourcing from this node.
"""Get physical replication slots configuration for a given member.
If the ``replicatefrom`` tag is set on the member - we should not create the replication slot for it on
the current primary, because that member would replicate from elsewhere. We still create the slot if
the ``replicatefrom`` destination member is currently not a member of the cluster (fallback to the
primary), or if ``replicatefrom`` destination member happens to be the current primary.
There are following situations possible:
If the ``nostream`` tag is set on the member - we should not create the replication slot for it on
the current primary or any other member even if ``replicatefrom`` is set, because ``nostream`` disables
WAL streaming.
* If the ``nostream`` tag is set on the member - we should not have the replication slot for it
on the current primary or any other member even if ``replicatefrom`` is set, because
``nostream`` disables WAL streaming.
* PostgreSQL is 11 and newer and configuration allows retention of member replication slots. In this case
we want to have replication slots for every member except the case when we have ``nofailover`` tag set.
* PostgreSQL is older than 11 or configuration doesn't allow member slots retention. In this case we want:
* On primary have replication slots for all members that don't have ``replicatefrom`` tag pointing
to the existing member.
* On replica node have replication slots only for members which ``replicatefrom`` tag pointing to us.
Will log an error if:
@@ -1195,7 +1202,7 @@ class Cluster(NamedTuple('Cluster',
ret[slot_name] = {'type': 'physical', 'lsn': lsn, 'expected_active': expected_active(member)}
slot_name = slot_name_from_member_name(name)
ret.update({slot: {'type': 'physical'} for slot in self.status.retain_slots
if slot not in ret and slot != slot_name})
if not nofailover and slot not in ret and slot != slot_name})
if len(ret) < len(members):
# Find which names are conflicting for a nicer error message
tests/test_slots.py
+14
View File
@@ -328,3 +328,17 @@ class TestSlotsHandler(BaseTestPostgresql):
None, None, None)], Exception])), \
patch.object(SlotsHandler, 'drop_replication_slot', Mock(return_value=(False, False))):
self.s.sync_replication_slots(cluster, self.tags)
@patch.object(Postgresql, 'is_primary', Mock(return_value=False))
@patch.object(Postgresql, 'role', PropertyMock(return_value='replica'))
@patch.object(TestTags, 'tags', PropertyMock(return_value={'nofailover': True}))
def test_slots_nofailover_tag(self):
self.p.name = self.leadermem.name
cluster = Cluster(True, ClusterConfig(1, {}, 1), self.leader,
Status(0, {}, [self.leadermem.name, self.other.name, self.me.name]),
[self.me, self.other, self.leadermem], None, SyncState.empty(), None, None)
global_config.update(cluster)
with patch.object(SlotsHandler, '_query', Mock(side_effect=[[('test_1', 'physical', 1, 12345, None, None,
None, None, None)], Exception])) as mock_query:
self.s.sync_replication_slots(cluster, self.tags)
self.assertTrue(mock_query.call_args[0][0].startswith('SELECT slot_name, slot_type, xmin, '))