Restore Postgres cluster after slave promotion

Luthfi 2017-11-20 22:56:33 UTC #1

Which version of the application are you using?:
Postgresql with Replication (9.6.5)
Please choose how you got the application: Installer (Windows, Linux, macOS), cloud image (AWS, GCE, Azure, ...) or VM (VMDK, VBOX):
Azure Marketplace
Have you installed any plugin or modified any configuration file?:
No
Describe here your question/suggestion/issue (expected and actual results):
I've deployed the Postgresql with replication from Bitnami on Azure, it has been running for a month when suddenly my web application cannot connect to the master database, I can connect locally from the terminal and remotely from pgAdmin. When I restart the master server from Azure portal the bitnami service reports a failure and the postgres service also cannot start. I have then promote the slave server into new master and have my web application to connect to the new master.

The question is, I want to retain this new master and just create a new slave db, how can I do that? For information the database has hundreds of schema and pg_dump command for all schema will run for hours before finally failed with out of shared memory error.

jsalmeron 2017-11-21 09:51:07 UTC #2

Hi,

If there are a lot of schema, could you just execute the pg_dump for only a subset of schemas? This may workaround the out of memory issue.

https://www.postgresql.org/docs/9.1/static/app-pgdump.html

I see here that there is an option --schema in the command line reference.

Best regards,

Javier J. Salmerón

Was my answer helpful? Click on

Luthfi 2017-11-21 12:08:25 UTC #3

Yes using per schema backup we can shorten the time and resolve the out of memory problem, I guess my real question is how can I turn the old master into the new slave, is there any guide for this? Especially for the current bitnami deployment.

silvio 2017-11-22 14:56:57 UTC #4

Hi,

From PostgreSQL 9.5 and on, postgreSQL includes a tool called pg_rewind which purpose is exactly what you need, turn old master in new slave in replication scenarios. In this case you can do:

pg_rewind --target-pgdata=/opt/bitnami/postgresql/data} --source-server="host=YOUR_NEW_MASTER_IP user=postgres dbname=postgres"

For more info and a deeper understanding of the tool please refer to these links:
Introduction to pg_rewind
pg_rewind in PG 9.5
pg_rewind in PG 9.6
pg_rewind official reference for PG 10

Regards,
Silvio Fernández

system 2017-12-06 14:57:00 UTC #5

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.

Bitnami provides free all-in-one installers, virtual machines and cloud images for popular open source applications. Get them now!