Grzegorz Uriasz

My metamask on firefox was also affected. The chrome version works fine.

@Ashniu123 Could you take a look at #210 ?

@cytadela8 Epoch stores the contract storage in the same tree(https://github.com/aeternity/protocol/blob/epoch-v0.16.0/serializations.md#contract) as all contracts.

CC: @velzevur @cytadela8

@kien-truong You can mitigate this by building a custom pooler image with `server_fast_close = 1`, this will ensure that when the underlying endpoint changes the DB connections will be reestablished....

@kien-truong I run PgBouncer in production in session mode and I've validated that server_fast_close helps, especially with long lived DB sessions. PgBouncer connects to the master postgres instance using the...

@kien-truong perhaps setting `dns_zone_check_period` will help?

Turns out that pgbouncer is pointed to a ClusterIP service. PGBouncer is unable to detect a failover of the master instance as it's always pointed to the same IP.

> > Also, though less important, Why master service is not using selector to choose the master pod ? > > Because Patroni manages the endpoint and therefore can guaranty...

@CyberDem0n I agree that races are unavoidable and there will be a period of time after a failover that connections to the postgres master are routed to a replica(due to...