PgBouncer
- Canonical
- Databases
Channel | Revision | Published | Runs on |
---|---|---|---|
latest/stable | 5 | 17 Jan 2022 | |
1/stable | 397 | 11 Sep 2024 | |
1/stable | 396 | 11 Sep 2024 | |
1/stable | 395 | 11 Sep 2024 | |
1/stable | 394 | 11 Sep 2024 | |
1/candidate | 397 | 02 Sep 2024 | |
1/candidate | 396 | 02 Sep 2024 | |
1/candidate | 395 | 02 Sep 2024 | |
1/candidate | 394 | 02 Sep 2024 | |
1/beta | 397 | 29 Aug 2024 | |
1/beta | 396 | 29 Aug 2024 | |
1/beta | 395 | 29 Aug 2024 | |
1/beta | 394 | 29 Aug 2024 | |
1/edge | 530 | Yesterday | |
1/edge | 529 | Yesterday | |
1/edge | 528 | Yesterday | |
1/edge | 527 | Yesterday |
juju deploy pgbouncer --channel 1/edge
Deploy universal operators easily with Juju, the Universal Operator Lifecycle Manager.
Platform:
Optimal PgBouncer Setup
For optimal performance, it is recommended that PgBouncer is run alongside your application. Co-locating PgBouncer with your application results in increased network performance since an additional network hop from your application to PgBouncer is avoided. Furthermore, it can also lead increased security since traffic is not routed externally through potentially untrusted machines.
When your application implements the modern (preferred) interface in PgBouncer’s supported interfaces , the PgBouncer charm is deployed as a subordinate of your application charm and your application.
Accessing PgBouncer outside of Juju
A known limitation of relating with PgBouncer (a subordinate charm) is that your application would need to be deployed as a Juju application. However, if your application exists outside of the Juju ecosystem, you can access PgBouncer externally with the Data Integrator charm.
Example setup
The steps below show you how to deploy and set up PostgreSQL, PgBouncer, and Data Integrator for access outside of Juju.
First, deploy all the charms:
juju deploy pgbouncer --channel 14/edge --trust
juju deploy data-integrator --config database-name=test_database
juju deploy pgbouncer --channel 1/edge
Feel free to change
test_database
to your name of choice
Integrate:
postgresql
withpgbouncer
data-integrator
withpgbouncer
, since in this case we want to generate the credentials to access PgBouncer
juju integrate postgresql pgbouncer
juju integrate data-integrator pgbouncer
The following is a sample output of the get-credentials
action run on a data-integrator
unit:
juju run data-integrator/leader get-credentials
ok: "True"
postgresql:
data: '{"database": "test_database", "external-node-connectivity": "true", "requested-secrets":
"[\"username\", \"password\", \"tls\", \"tls-ca\", \"uris\"]"}'
database: test_database
endpoints: 10.205.193.65:6432
password: mysupersecuredatabasepassword
subordinated: "true"
uris: postgresql://relation_id_7:mysupersecuredatabasepassword@10.205.193.65:6432/test_database
username: relation_id_7
version: "14.13"
You can then connect to PgBouncer with the provided uris
from your application that resides outside of Juju.
Using a Virtual IP to connect to PgBouncer
If the PgBouncer charm is related to the Data Integrator charm, it is possible for a user to configure PgBouncer to use a certain Virtual IP. This assumes that the user has somehow ensured that the Virtual IP resolves to the node on which the PgBouncer charm is deployed.
To configure the PgBouncer charm with a virtual IP, run
juju config pgbouncer/0 vip=<your_virtual_ip>
Integrate with HACluster
Alternatively, you can integrate with the HACluster charm if you would like a Virtual IP that is generated and maintained for you.
HACluster is a collection of solutions by ClusterLabs designed to create and manage resources. The creation of resources like Virtual IPs is handled by Pacemaker, whereas the management of these resources is handled by Corosync.
Pacemaker will create and attach a Virtual IP to one of your Data Integrator nodes (that is related to PgBouncer), while Corosync will ensure automatic failover if the node with the Virtual IP faces connectivity or other issues. This setup requires at least 3 Data Integrator nodes, each related to both PgBouncer and HACluster.
Warning: The Virtual IP supplied to PgBouncer should be in the same subnet as the nodes on which the PgBouncer charm is running. Else, you may encounter unexpected behavior from the HACluster charm when it tries to create the Virtual IP.
Example setup
The steps below show you how to deploy and set up PostgreSQL, PgBouncer, Data Integrator, and HACluster.
First, deploy all the charms
juju deploy postgresql --channel 14/edge --trust
juju deploy -n 3 data-integrator --config database-name=test_database
juju deploy pgbouncer --channel dpe/edge
juju deploy hacluster
Note that the
data-integrator
requires a minimum of 3 nodes for this HACluster setup to work
Configure the VIP on pgbouncer
. Please ensure that the VIP is in an accessible subnet:
juju config pgbouncer vip=10.205.193.35
Then, integrate:
pgbouncer
withpostgresql
pgbouncer
andhacluster
withdata-integrator
pgbouncer
withhacluster
juju integrate pgbouncer postgresql
juju integrate data-integrator pgbouncer
juju integrate data-integrator:juju-info hacluster:juju-info
juju integrate pgbouncer hacluster
The following is a sample output of the get-credentials
action run on a data-integrator
unit:
juju run data-integrator/leader get-credentials
Running operation 3 with 1 task
- task 4 on unit-data-integrator-0
Waiting for task 4...
ok: "True"
postgresql:
data: '{"database": "test_database", "external-node-connectivity": "true", "requested-secrets":
"[\"username\", \"password\", \"tls\", \"tls-ca\", \"uris\"]"}'
database: test_database
endpoints: 10.205.193.35:6432
password: mysupersecuredatabasepassword
read-only-endpoints: 10.205.193.80:6432,10.205.193.98:6432
subordinated: "true"
uris: postgresql://relation_id_7:mysupersecuredatabasepassword@10.205.193.35:6432/test_database
username: relation_id_7
version: "14.13"