Closed benmccown closed 5 years ago
Note that this is with the latest version of CentOS (7.6)
I have have committed a solution to my fork and will create a pull request shortly after some more testing
It looks like there's more needed to have PG 10 fully operational. At least, we need a new version of PAF.
I've pushed a new PR #21 to cover that. But there's an issue in PAF module itself with checking of notify
parameter. The status can be seen here: ClusterLabs/PAF#141
Jan
Sounds good. I have run out of time today but hope to test Postgres 10 more tomorrow (Friday) and through next week.
Thanks, Ben
Done.
I've created a PR to ClusterLabs/PAF and meanwhile I've created a patch that is applied during the PAF install in this role. Tested and working with PG versions 9.6, 10, 11. Happy ansibling.
Jan
PS: together with this, I've pushed my additional optional PAF patch that helps to better handle cluster network splits. It can be enabled by postgres_ha_paf_geo_patch
variable. More info here.
I'm trying to use this role to deploy Postgres 10 after using it for quite a few 9.6 clusters over the last couple months.
The postgresql_sync.yml play fails on the "init DB dir on master if necessary" due to a change in the filename for postgres version 10 (and I'm guessing above.)
In the bin dir for postgres 9.6:
Notice the "postgresql96-setup" script
Now in postgres 10's bin dir:
Notice the added dash after "postgresql" in "postgresql-10-setup"
This seems to cause some obvious problems with the playbook.