Consider a situation: there is a permanent logical slot and primary and replica are temporary down.
When Patroni is started on the former primary it starts Postgres in a standby mode, what leads to removal of physical replication slot for the replica because it has xmin.
We should postpone removal of such physical slots:
- on replica until there will be a leader in the cluster
- on primary until Postgres is promoted
Since `3.2.0` Patroni is able to create physical replication slots on replica nodes just for the case if this node at some moment will become the primary.
There are two potential problems of having such slots:
1. They prevent recycling of WAL files.
2. They may affect vacuum on the primary is hot_standby_feedback is enabled.
The first class of issues is already addressed by periodically calling pg_replication_slot_advance() function.
However the second class of issues doesn't happen instantly, but only when the old primary switched to a replica. In this case physical replication slots that were at some moment activate will hold NOT NULL value of `xmin`, which will be propagated to the primary via hot_standby_feedback mechanism.
To address the second problem we will detect that a physical replication slot is not supposed to be active, but having NOT NULL `xmin` and drop/crecreate it.
Close#3146Close#3153
Co-authored-by: Polina Bungina <[email protected]>
Current problem of Patroni that strikes many people is that it removes replication slot for member which key is expired from DCS. As a result, when the replica comes back from a scheduled maintenance WAL segments could be already absent, and it can't continue streaming without pulling files from archive.
With PostgreSQL 16 and newer we get another problem: logical slot on a standby node could be invalidated if physical replication slot on the primary was removed (and `pg_catalog` vacuumed).
The most problematic environment is Kubernetes, where slot is removed nearly instantly when member Pod is deleted.
So far, one of the recommended solutions was to configure permanent physical slots with names that match member names to avoid removal of replication slots. It works, but depending on environment might be non-trivial to implement (when for example members may change their names).
This PR implements support of `member_slots_ttl` global configuration parameter, that controls for how long member replication slots should be kept when the member key is absent. Default value is set to `30min`.
The feature is supported only starting from PostgreSQL 11 and newer, because we want to retain slots not only on the leader node, but on all nodes that could potentially become the new leader, and they should be moved forward using `pg_replication_slot_advance()` function.
One could disable feature and get back to the old behavior by setting `member_slots_ttl` to `0`.
The `Status` class was introduced in #2853, but we kept old properties in the `Cluster` object in order to have fewer changes in the rest of the code.
This PR is finishing the refactoring.
The following adjustments were made:
- Introduced `Status.is_empty()` method, which is used in the `Cluster.is_empty()` instead of checking actual values to simplify introduction of further fields to the Status object.
- Removed `Cluster.last_lsn` property
- Changed `Cluster.slots` property to always return dict and perform sanity checks on values.
Besides that, this PR addressing a couple of problems:
- the `AbstractDCS.get_cluster()` method some properties without holding a lock on `_cluster_thread_lock`.
- `Cluster.__permanent_slots` property was setting 'lsn' from all cluster members, while it should be doing that only for members with `replicatefrom` tag.
The `SlotsAdvanceThread` is asynchronously calling
pg_replication_slot_advance() and providing feedback about logical
replication slots that must be reinitialized by copying from the
primary. That is, the parent thread will learn about slots to be copied
only when scheduling the next pg_replication_slot_advance() call.
As a result it was possible situation when logical slot was copied with
PostgreSQL restart more than once.
To improve it we implement following measures:
1. do not schedule slot sync if it is in the list to be copied
2. remove to be copued slots from the `self._scheduled` structure
3. clean state of `SlotsAdvanceThread` when slot files are copied.
Since PG16 logical replication slots on a standby can be invalidated due
to horizon. In this case, pg_replication_slot_advance() will fail with
ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE. We should force slot copy
(i.e., recreation) of such slots.
Besides that:
1. fix problem with is_physical_slot() methods, it was returning false positives for logical slots.
2. Fix a little issue with replicatefrom docs.
Close https://github.com/zalando/patroni/issues/3068
Add support for ``nostream`` tag. If set to ``true`` the node will not use replication protocol to stream WAL. It will rely instead on archive recovery (if ``restore_command`` is configured) and ``pg_wal``/``pg_xlog`` polling. It also disables copying and synchronization of permanent logical replication slots on the node itself and all its cascading replicas. Setting this tag on primary node has no effect.
Instead of passing around names, specific tags, and Postgres version just pass Postgresql object and objects implementing Tags interface.
It should simplify implementation of #2842
1. extract `GlobalConfig` class to its own module
2. make the module instantiate the `GlobalConfig` object on load and replace sys.modules with the this instance
3. don't pass `GlobalConfig` object around, but use `patroni.global_config` module everywhere.
4. move `ignore_slots_matchers`, `max_timelines_history`, and `permanent_slots` from `ClusterConfig` to `GlobalConfig`.
5. add `use_slots` property to global_config and remove duplicated code from `Cluster` and `Postgresql.ConfigHandler`.
Besides that improve readability of couple of checks in ha.py and formatting of `/config` key when saved from patronictl.
- Fixed issues with has_permanent_slots() method. It didn't took into account the case of permanent physical slots for members, falsely concluding that there are no permanent slots.
- Write to the status key only LSNs for permanent slots (not just for slots that exist on the primary).
- Include pg_current_wal_flush_lsn() to slots feedback, so that slots on standby nodes could be advanced
- Improved behave tests:
- Verify that permanent slots are properly created on standby nodes
- Verify that permanent slots are properly advanced, including DCS failsafe mode
- Verify that only permanent slots are written to the `/status`
Create permanent physical replication slots on standby nodes and use `pg_replication_slot_advance()` function to move them forward.
The `restart_lsn` is advanced based on values stored in the `/status` key by the primary node.
When slot is created on a replica it could be ahead the same slot on the primary and therefore there is some period of time when it doesn't protect WAL files from being recycled.
It represents the `/status` key in DCS and makes it easier to introduce new values stored in the `/status` key without need to refactor all DCS implementations.
1. make _get_members_slots() method return data in the same format as _get_permanent_slots() method
2. move conflicting name handling from get_replication_slots() to _get_members_slots() method
3. enrich structure returned by get_replication_slots() with the LSN of permanent logical slots reported by primary
4. use the added information in the SlotsHandler instead of fetching it from the Cluster.slots
5. bugfix: don't try to advance logical slot that doesn't match required configuration
Cluster.get_replication_slots() didn't take into account that there can not be logical replication slots in a standby cluster replicas. It was only skipping logical slots for the standby_leader, but replicas were expecting that they will have to copy them over.
Also on replicas in a standby cluster these logical slots were falsely added to the `_replication_slots` dict.
1. stop using the same cursor all the time, it creates problems when not carefully used from different threads.
2. introduce query() method in the Connection class and make it return a result set when it is possible.
3. refactor most of the code that is relying (directly or indirectly) on the Connection object to use the query() method as much as possible.
This refactoring helps with reducing code complexity and will help with future introduction of a separate database connection for the REST API thread. The last one will help to improve reliability when system is under significant stress when simple monitoring queries are taking seconds to execute and the REST API starts blocking the main thread.
Includes:
* renaming of `_unready_logical_slots` to better represent that it is a processing queue which is emptied on successful completion.
* ensuring that return type is consistent.
* made logic variable names explicit to help explain how the decision of whether a slot is "ready" is made.
To do that we use `pg_stat_get_wal_receiver()` function, which is available since 9.6. For older versions the `patronictl list` output and REST API responses remain as before.
In case if there is no wal receiver process we check if `restore_command` is set and show the state as `in archive recovery`.
Example of `patronictl list` output:
```bash
$ patronictl list
+ Cluster: batman -------------+---------+---------------------+----+-----------+
| Member | Host | Role | State | TL | Lag in MB |
+-------------+----------------+---------+---------------------+----+-----------+
| postgresql0 | 127.0.0.1:5432 | Leader | running | 12 | |
| postgresql1 | 127.0.0.1:5433 | Replica | in archive recovery | 12 | 0 |
+-------------+----------------+---------+---------------------+----+-----------+
$ patronictl list
+ Cluster: batman -------------+---------+-----------+----+-----------+
| Member | Host | Role | State | TL | Lag in MB |
+-------------+----------------+---------+-----------+----+-----------+
| postgresql0 | 127.0.0.1:5432 | Leader | running | 12 | |
| postgresql1 | 127.0.0.1:5433 | Replica | streaming | 12 | 0 |
+-------------+----------------+---------+-----------+----+-----------+
```
Example of REST API response:
```bash
$ curl -s localhost:8009 | jq .
{
"state": "running",
"postmaster_start_time": "2023-07-06 13:12:00.595118+02:00",
"role": "replica",
"server_version": 150003,
"xlog": {
"received_location": 335544480,
"replayed_location": 335544480,
"replayed_timestamp": null,
"paused": false
},
"timeline": 12,
"replication_state": "in archive recovery",
"dcs_last_seen": 1688642069,
"database_system_identifier": "7252327498286490579",
"patroni": {
"version": "3.0.3",
"scope": "batman"
}
}
$ curl -s localhost:8009 | jq .
{
"state": "running",
"postmaster_start_time": "2023-07-06 13:12:00.595118+02:00",
"role": "replica",
"server_version": 150003,
"xlog": {
"received_location": 335544816,
"replayed_location": 335544816,
"replayed_timestamp": null,
"paused": false
},
"timeline": 12,
"replication_state": "streaming",
"dcs_last_seen": 1688642089,
"database_system_identifier": "7252327498286490579",
"patroni": {
"version": "3.0.3",
"scope": "batman"
}
}
```
- added pyrightconfig.json with typeCheckingMode=strict
- added type hints to all files except api.py
- added type stubs for dns, etcd, consul, kazoo, pysyncobj and other modules
- added type stubs for psycopg2 and urllib3 with some little fixes
- fixes most of the issues reported by pyright
- remaining issues will be addressed later, along with enabling CI linting task
It was possible to have it empty if the all cluster keys are missing in DCS. In this case the `Cluster` object was manually created with all values set to `None` or `[]` (including sync).
It already resulted in #2217, which is in fact wasn't a correct fix.
In order to solve it and reduce code duplication we introduce `Cluster.empty()` and `SyncState.empty()` methods, which will create corresponding empty objects and start using `Cluster.empty()` from all places where the empty `Cluster` object was manually created.
keep as much backward compatibility as possible.
Following changes were made:
1. All internal checks are performed as `role in ('master', 'primary')`
2. All internal variables/functions/methods are renamed
3. `GET /metrics` endpoint returns `patroni_primary` in addition to `patroni_master`.
4. Logs are changed to use leader/primary/member/remote depending on the context
5. Unit-tests are using only role = 'primary' instead of 'master' to verify that 1 works.
6. patronictl still supports old syntax, but also accepts `--leader` and `--primary`.
7. `master_(start|stop)_timeout` is automatically translated to `primary_(start|stop)_timeout` if the last one is not set.
8. updated the documentation and some examples
Future plan: in the next major release switch role name from `master` to `primary` and maybe drop `master` altogether.
The Kubernetes implementation will require more work and keep two labels in parallel. Label values should probably be configurable as described in https://github.com/zalando/patroni/issues/2495.
On busy clusters with many logical replication slots the pg_replication_slot_advance () call affects the main HA loop and could result in the member key expiration.
The only way to solve it is a dedicated thread response for moving slots forward.
The thread is started only when there are logical slots to be advanced.
Will help to solve #2388Close#2239
If replication slots are enabled Patroni automatically creates them for any cluster member that is supposed to stream from a given node and for any permanent slot defined in the global configuration. If the member disappears from the DCS Patroni automatically removes the replication slot for it. The same behavior was in the maintenance mode (pause).
This commit disables removal of any replication slots that don't match Patroni's expectations in pause.
Close https://github.com/zalando/patroni/issues/2314
The logical slot on a replica is safe to use when the physical replica
slot on the primary:
1. has a nonzero/non-null `catalog_xmin`
2. has a `catalog_xmin` that is not newer (greater) than the `catalog_xmin` of any slot on the standby
3. the `catalog_xmin` is known to overtake `catalog_xmin` of logical slots on the primary observed during `1`
In case if `1` doesn't take place, Patroni will run an additional check whether the `hot_standby_feedback` is actually in effect and shows the warning in case it is not.
A couple of times we have seen in the wild that the database for the permanent logical slots was changed in the Patroni config.
It resulted in the below situation.
On the primary:
1. The slot must be dropped before creating it in a different DB.
2. Patroni fails to drop it because the slot is in use.
Replica:
1. Patroni notice that the slot exists in the wrong DB and successfully dropping it.
2. Patroni copying the existing slot from the primary by its name with Postgres restart.
And the loop repeats while the "wrong" slot exists on the primary.
Basically, replicas are continuously restarting, which badly affects availability.
In order to solve the problem, we will perform additional checks while copying replication slot files from the primary and discard them if `slot_type`, `database`, or `plugin` don't match our expectations.
It could happen that the replica for some reason is missing the WAL file required by the replication slot.
The nature of this phenomenon is a bit unclear, it might be that the WAL was recycled short before we copied the slot file, but, we still need a solution to this problem. If the `pg_replication_slot_advance()` fails with the `UndefinedFile` exception (requested WAL segment pg_wal/... has already been removed), the logical slot on the replica must be recreated.
Effectively, this PR consists of a few changes:
1. The easy part:
In case of permanent logical slots are defined in the global configuration, Patroni on the primary will not only create them, but also periodically update DCS with the current values of `confirmed_flush_lsn` for all these slots.
In order to reduce the number of interactions with DCS the new `/status` key was introduced. It will contain the json object with `optime` and `slots` keys. For backward compatibility the `/optime/leader` will be updated if there are members with old Patroni in the cluster.
2. The tricky part:
On replicas that are eligible for a failover, Patroni creates the logical replication slot by copying the slot file from the primary and restarting the replica. In order to copy the slot file Patroni opens a connection to the primary with `rewind` or `superuser` credentials and calls `pg_read_binary_file()` function.
When the logical slot already exists on the replica Patroni periodically calls `pg_replication_slot_advance()` function, which allows moving the slot forward.
3. Additional requirements:
In order to ensure that primary doesn't cleanup tuples from pg_catalog that are required for logical decoding, Patroni enables `hot_standby_feedback` on replicas with logical slots and on cascading replicas if they are used for streaming by replicas with logical slots.
4. When logical slots are copied from to the replica there is a timeframe when it could be not safe to use them after promotion. Right now there is no protection from promoting such a replica. But, Patroni will show the warning with names of the slots that might be not safe to use.
Compatibility.
The `pg_replication_slot_advance()` function is only available starting from PostgreSQL 11. For older Postgres versions Patroni will refuse to create the logical slot on the primary.
The old "permanent slots" feature, which creates logical slots right after promotion and before allowing connections, was removed.
Close: https://github.com/zalando/patroni/issues/1749
