1. Fix race condition on shutdown. It is very annoying when you cancel behave tests but postgres remains running.
2. Dump pg_controldata output to logs when "recovering" stopped postgres. It will help to investigate some annoying issues.
First of all, this patch changes the behavior of `on_start`/`on_restart` callbacks, they will be called only when postgres is started or restarted without role changes. In case if the member is promoted or demoted only the `on_role_change` callback will be executed. `on_role_change` was never called for standby leader, only `on_start`/`on_restart` and with a wrong role argument.
Before that `on_role_change` was never called for standby leader, only `on_start`/`on_restart` and with a wrong role argument.
In addition to that, the REST API will return standby_leader role for the leader of the standby cluster.
Closes https://github.com/zalando/patroni/issues/988
`dcs.cluster` and `dcs.get_cluster()` are using the same lock resource and therefore when get_cluster call is slow due to the slowness of DCS it was also affecting the `dcs.cluster` call, which in return was making health-check requests slow.
In addition to that transfer postmaster pid to Patroni process with the help of multiprocessing.Pipe instead of using stdin-stdout pipes.
Closes https://github.com/zalando/patroni/issues/992
if the `etcd.use_proxies` is set to true, Patroni will stick to the list of hosts specified in the `etcd.hosts` and avoid doing topology discovery. Such mode might be useful when you know that you connect to the etcd cluster via the set of proxies or when th etcd cluster has static topology.
1. Multi-stage build with an extensive cleanup of useless files and optional image compression
2. Start three-node etcd cluster
3. Start three-node Patroni cluster
4. One container with haproxy
5. All container names are prefixed with "demo-" and don't have suffixes
6. Decommission dev_patroni_cluster.sh script, docker-compose is now standard de-facto.
7. Provide more examples in the docker/README.md
It might happen that the standby cluster is configured to be created and replay WAL files from the different source than when it is running not in standby mode. This is necessary to avoid writing WAL files and backups into the old place after promotion.
The easiest way to achieve such behavior is passing RemoteMember object to `Postgresql.clone` method instead of the usual Member object.
if there is no service defined k8s assumes that endpoint is orphaned and removes it.
Patroni tries to create the service only in case if use_endpoints is enabled if the following cases:
1. Upon start
2. When it tries to (re-)create the config endpoint
If for some reason creation of the service has failed, Patroni will retry it on every cycle of HA loop. Usually it fails due to lack of permissions and if you don't want to give such permissions to the service account used by Patroni, you can create the service explicitly in the deployment manifest.
According to the Consul documentation the actual response timeout is increased by a small random amount of additional wait time added to the supplied maximum wait time to spread out the wake up time of any concurrent requests. It adds up to wait / 16 additional time to the maximum duration.
In our case we will add wait/15 or 1 second depending on what is bigger.
Fixes: https://github.com/zalando/patroni/issues/945
This information will help to detect stale replicas.
In addition to that Host will include ':{port}' if the port value isn't default or more than one member running on the same host.
Fixes: https://github.com/zalando/patroni/issues/942
If the pg_rewind is disabled or can't be used, the former master could fail to start as a new replica due to diverged timelines. In this case, the only way to fix it is wiping the data directory and reinitializing.
So far Patroni was able to remove the data directory only after failed attempt to run pg_rewind. This commit fixes it.
If the `postgresql.remove_data_directory_on_diverged_timelines` is set, Patroni will wipe the data directory and reinitialize the former master automatically.
Fixes: https://github.com/zalando/patroni/issues/941
The latest timeline is calculated from the `/history` key in DCS. In case there is no such key or it contains some garbage we consider the node healthy.
Closes https://github.com/zalando/patroni/issues/890
libpq allows opening connection without explicitly specifying neither username nor password. Depending on situation it would rely either on `pgpass` file or trust authentication method in pg_hba.conf.
Since pg_rewind is also using libpq, it could work the same way.
Fixes https://github.com/zalando/patroni/issues/928
It allows changing logging settings in runtime by updating config and doing reload or sending `SIGHUP` to the Patroni process.
Important! Environment configuration names related to logging were renamed and documentation accordingly updated. For compatibility reasons Patroni still accepts `PATRONI_LOGLEVEL` and `PATRONI_FORMAT`, but some other variables related to logging, which were introduced only
recently (between releases), will stop working. I think it is ok, since we didn't release the new version yet and therefore it is very unlikely that somebody is using them except authors of corresponding PRs.
Example of log section in the config file:
```yaml
log:
dir: /where/to/write/patroni/logs # if not specified, write logs to stderr
file_size: 50000000 # 50MB
file_num: 10 # keep history of 10 files
dateformat: '%Y-%m-%d %H:%M:%S'
loggers: # increase log verbosity for etcd.client and urllib3
etcd.client: DEBUG
urllib3: DEBUG
```
1. Log only debug level messages on any kind of error
2. Update regexp for matching postgres aux processes to make it compatible with postgres 11
Fixes https://github.com/zalando/patroni/issues/914
We want to avoid archiving WALs and history files until the cluster is fully functional. It should really help if the custom bootstrap involves pg_upgrade.
Change of `loop_wait` was causing Patroni to disconnect from zookeeper and never reconnect back. The error was happening only with python3 due to a difference in implementation of `select.select` function.
Recently 2.6.0 was release which changes the way how create_connection method is called. Before it was passing two arguments, and in the new version all argument names are specified explicitly.
Permanent replication slots are preserved on failover/switchover, that is Patroni on the new primary will create configured replication slots right after doing promote.
Slots could be configured with the help of `patronictl edit-config`.
The initial configuration could be also done in the `bootstrap.dcs`
```yaml
slots:
permanent_physical_1:
type: physical
permanent_logical_1:
type: logical
database: foo
plugin: pgoutput
```
It is the responsibility of the operator to make sure that there are no clashes in names between replication slots automatically created by Patroni for members and permanent replication slots.
Closes https://github.com/zalando/patroni/issues/656
* Always run `pg_rewind` against the remote master
* Always use the remote master as the source when "recovering" stopped standby leader
* Use remote master as the source when "recovering" the node in the unhealthy cluster
* Use the local dbname as the fallback when doing `pg_rewind` from the remote master
* `no_replication_slot` is the allowed key in the `RemoteMember` object
* Make it possible to "bootstrap" the new `standby_cluster` with existing (and valid) data directory. There is one prerequisite though, there should be no `patroni.dynamic.json` file in it!
* Use `shutil.move` instead of `os.replace`, which is available only from 3.3
* Introduce standby-leader health-check and consul service
* Improve unit tests, some lines were not covered
* rename `assertEquals` -> `assertEqual`, due to deprecation warning
In synchronos_mode_strict we put '*' into synchronos_standby_names, what makes one connection 'sync' and other connections 'potential'.
The code picking up the correct sync standby didn't consider 'potential' as a good candidate.
Fixes: https://github.com/zalando/patroni/issues/789
If Patroni gets partitioned it starts receiving stale information from DCS.
We can't use this information to determine that we have the leader key.
Instead, we will record in Ha object the actual state of acquire/update lock and report as a leader only if it was successful.
P.S. despite responding with 200 on `GET /master` postgres was still running read-only.
In really rare cases it was causing following behavior:
```
2018-07-31 10:35:30,302 INFO: starting as a secondary
2018-07-31 10:35:30,309 INFO: Lock owner: postgresql0; I am postgresql1
2018-07-31 10:35:30,310 INFO: Demoting master during restarting after failure
2018-07-31 10:35:30,381 INFO: postmaster pid=17709
2018-07-31 10:35:30,386 INFO: lost leader lock during restarting after failure
2018-07-31 10:35:30,388 ERROR: Exception during CHECKPOINT
```
Patroni is relying on params to determinte timeout and amount of retries when executing api requests to consul. Starting from v1.1.0 python-consul changed internal API and started
using `list` instead of `dict` to pass query parameters. Such change broke "watch" functionality.
Fixes https://github.com/zalando/patroni/issues/742 and
https://github.com/zalando/patroni/issues/734
* Take and apply some parameters from controldata when starting as replica
https://www.postgresql.org/docs/10/static/hot-standby.html#HOT-STANDBY-ADMIN
There is set of parameters which value on the replica must be not smaller than on the primary, otherwise replica will refuse to start:
* max_connections
* max_prepared_transactions
* max_locks_per_transaction
* max_worker_processes
It might happen that values of these parameters in the global configuration are not set high enough, what makes impossible to start a replica without human intervention. Usually it happens when we bootstrap a new cluster from the basebackup.
As a solution to this problem we will take values of above parameters from the pg_controldata output and in case if the values in the global configuration are not high enough, apply values taken from pg_controldata and set `pending_restart` flag.
We already have a lot of logic in place to prevent failover in such case and restore all keys, but an accidental removal of `/config` key was effectively switching off pause mode for 1 cycle of HA loop.
Upon start postmaster process performs various safety checks if there is a postmaster.pid file in the data directory. Although Patroni already detected that the running process corresponding to the postmaster.pid is not a postmaster, the new postmaster might fail to start, because it thinks that postmaster.pid is already locked.
Important!!! Unlink of postmaster.pid isn't an option in this case, because it has a lot of nasty race conditions.
Luckily there is a workaround to this problem, we can pass the pid from postmaster.pid in the `PG_GRANDPARENT_PID` environment variable and postmaster will ignore it.
More likely to hit such problem if you run Patroni and postgres in the docker container.
It didn't affect directly neither failover nor switchover, but in some rare cases it was reporting it as a success too early, when the former leader released the lock: `Failed over to "None" instead of "desired-node"`
In addition to that this commit improves logs and status messages by differentiating between failover and switchover.
Patroni can attach itself to an already running PostgreSQL instance. If that is the first instance "seen" in the given cluster, Patroni for that instance will create the initialize key, grab the leader key and, if the instance is running a replica, promote.
Because of this behavior, when a cluster with a master and one or more replicas gets Patroni for each node, it is imperative to start running Patroni on the master node before getting to the replicas.
This commit changes such weird behavior and will abort Patroni start if there is no initialize key in DCS and postgres is running as a replica.
Closes https://github.com/zalando/patroni/issues/655
Because they are indeed case insensitive.
Most of the parameters have snake_case_name, but there are three exceptions from this rule: DateStyle, IntervalStyle and TimeZone.
In fact, if you specify timezone = 'some/tzn' it still works, but Patroni wasn't able to find 'timezone' in pg_settings and stripping this parameter out.
We will use CaseInsensitiveDict to keep postgresql.parameters. This change affects only "final" configuration. That means if you put some"duplicates" (work_mem vs WORK_MEM) into patroni yaml or into cluster config, it would be resolved only at the last stage and for example you will be able to see both values if you use `patronictl edit-config`.
Fixes https://github.com/zalando/patroni/issues/649
Output exception trace to the logs when http status code == 403, something is wrong with permissions.
When http status code == 409 -- such error could be ignored, because object probably was created or updated by another process.
For all other http status codes it will also produce stack traces.
I hope it will help to debug issues similar to the https://github.com/zalando/patroni/issues/606
It is very easy to get current timeline on the master by executing
```sql
SELECT ('x' || SUBSTR(pg_walfile_name(pg_current_wal_lsn()), 1, 8))::bit(32)::int
```
Unfortunately the same method doesn't work when postgres is_in_recovery. Therefore we will use replication connection for that on the replicas. In order to avoid opening and closing replication connection on every HA loop we will cache the result if its value matches with the timeline of the master.
Also this PR introduces a new key in DCS: `/history`. It will contain a json serialized object with timeline history in a format similar to the usual history files. The differences are:
* Second column is the absolute wal position in bytes, instead of LSN
* Optionally there might be a fourth column - timestamp, (mtime of history file)
Every iteration of HA loop Patroni needs to call pg_is_in_recovery() and calcualte absolute wal_position. It was doing two separate SELECT statements for that. In case of master it was doing even three queries (wal_position two times).
We will issue one SELECT for every HA loop and cache the results.
Make it possible to cancel a running task if you want to reinitialize replica.
There are two possible ways to trigger it:
1. patronictl will ask whether you want to cancel already running task if an attempt to trigger reinitialize has failed
2. if you are using `--force` argument with `patronictl reinit`
This list will be used for initial discovery of etcd cluster members.
If for some reason during work this list of hosts has been exhausted (during work), Patroni will return to initial list.
In addition to that improve ipv6 compatibility by using a special function for splitting host and port.
Fixes https://github.com/zalando/patroni/issues/523
* Use ConfigMaps or Endpoins for leader elections and to keep cluster state
* Label pods with a postgres role
* change behavior of pip install. From now on it will not install all dependencies, you have to specify explicitly DCS you want to use Patroni with: `pip install patroni[etcd,zookeeper,kubernetes]`
* Show information about scheduled failover and maintenance mode when showing list of cluster members. Fixes https://github.com/zalando/patroni/issues/557
* Fix postgres version check functions (postgres 10 and above compatibility) and apply pep8 formatting to the tests.
* Bump some configuration parameters to match with postgres 10 defaults.
* Fix name of contributor in release notes.
