[Question] Issues with pgdog not auto banning failed database

[SOG-web]

Am using repmgr for auto failover

in a case where the primary is stopped, I detected that repmgr auto select new primary

but pgdog stop responding due to postgres1 down

but since repmgr already selected a new primary is there a way to make pgdog ignore the down of postgres1, or what other solution do we have

pgdog.toml

[general]
host = "0.0.0.0"
port = 6432
shutdown_timeout = 1000
openmetrics_port = 9090
query_timeout = 30000
checkout_timeout = 10000
connect_timeout = 5000
idle_healthcheck_interval = 5000
ban_timeout = 60000

read_write_split = "include_primary_if_replica_banned"
read_write_strategy = "conservative"
load_balancing_strategy = "least_active_connections"

pooler_mode = "session"
default_pool_size = 20

lsn_check_delay = 0
lsn_check_interval = 1_000
ban_replica_lag = 60_000
ban_replica_lag_bytes = 25_000_000

[[databases]]
name = "postgres"
host = "postgres1"
port = 5432
role = "auto"


[[databases]]
name = "postgres"
host = "postgres2"
port = 5432
role = "auto"


[[databases]]
name = "postgres"
host = "postgres3"
port = 5432
role = "auto"


[tcp]
retries = 3
time = 1000
interval = 1000
user_timeout = 1000

[admin]
name = "admin"
user = "admin"
password = "PGDOG_ADMIN_PASSWORD_PLACEHOLDER"

users.toml

[[users]]
name = "postgres"
database = "postgres"
password = "POSTGRES_PASSWORD_PLACEHOLDER"
server_user = "postgres"
server_password = "POSTGRES_PASSWORD_PLACEHOLDER"

[[users]]
name = "app"
database = "postgres"
password = "APP_PASSWORD_PLACEHOLDER"
server_user = "app"
server_password = "APP_PASSWORD_PLACEHOLDER"

[levkk]

Hi. Is pgdog connecting directly to the databases or through another proxy? Also, which version of pgdog are you using?

[SOG-web]

latest version

It seems the database is connecting now through pgdog

so what happened was that I tried to connect before the ban happened

but if I may ask if the down database comes back online in the future will pgdog auto connect ?
also is there a way to allow dynamic database creation ?

but for now this is the solution am thinking of using

create a recuring function in the agent to check for when any db container is down/not reachable/unhealthy

then it should follow this rules

1. confirm the current primary from repmgr - keep running this until it gets a primary (circle through the db containers till one give us a proper answer)
2. edit the pgdog config and reload pgdog
3. remove the faulty and clean the volume and every necessary stuff regarding the down container
4. recreate it as a replica
5. edit the pgdog config relaod pgdog