Evaluating high availability solutions for TimescaleDB and PostgreSQL

[fanf2]

https://blog.timescale.com/high-availability-timescaledb-postgresql-patroni-a4572264a831

[adamfeldman]

I'm interested in solutions for pg HA that minimize data loss risk. Has anyone used Patroni's synchronous replication? How was your experience?

In that vein, this looks important to maximizing safety in the event of split brain: https://github.com/zalando/patroni/blob/master/docs/watchdog...

Patroni synchronous mode docs: https://github.com/zalando/patroni/blob/master/docs/replicat...

[ddorian43]

Any solution that also has sharding (excluding citus that is single master) ?

[mfreed → ddorian43]

Hi @ddorian43, Timescale person here. I assume the question is asking more about TimescaleDB, rather than an HA solution that is shard-aware in is View management?

We're in the process of building our own scale-out clustering solution, so that the many chunks that belong to a hypertable (some users have 10,000s of such chunks per hypertable) can be spread across many servers, rather than on a single primary as currently.

( Plan is to release this coming year; you can already see some PRs hit out public github about changes for this effect: https://github.com/timescale/timescaledb/pull/904 )

[lykr0n → mfreed]

That's awesome. I think this is one the major things holding back TimescaleDB. There are limits of how big you can reasonably make a Linux Server.

[ddorian43 → mfreed]

No I was asking about pg in general.