[HN Gopher] The Wonders of Postgres Logical Decoding Messages ___________________________________________________________________ The Wonders of Postgres Logical Decoding Messages Author : rbanffy Score : 63 points Date : 2023-03-31 20:21 UTC (1 days ago) (HTM) web link (www.infoq.com) (TXT) w3m dump (www.infoq.com) | cryptonector wrote: | NOTIFY also can be used this way. | klysm wrote: | Notify is really hard to use correctly, and I think logical | replication works better in most cases | semiquaver wrote: | Correct me if I'm wrong but even though NOTIFY is | transactional, it isn't persisted to WAL or anywhere else --- | so any clients that were not LISTENing at the time NOTIFY was | called cannot receive it retroactively. WAL is, well, a _log_ | which has nicer properties for many applications that allow | clients to "fall behind" or recover from transient network | disconnects without losing data. | aynyc wrote: | Does it support some form of checkpointing without Flink? I would | like to run debezium without Flink. | jakehilborn wrote: | Yes. Postgres stores the checkpoint on the source database | replication slot. You can configure Debezium Engine to send | checkpoints to the file /dev/null to disable the default | external system checkpoint behavior. | aynyc wrote: | Any documentation on this? We ran an assortment of outbox | pattern for data distribution, I don't want to manage that in | data producers anymore. | simonw wrote: | I remember hearing years ago that Facebook used a trick like this | for cache invalidation: they would write custom "invalidate cache | X" messages (maybe even as SQL comments?) into their MySQL | replication streams that fanned out to multiple data centers, | such that a data center would know to clear a cached record out | of its local memcached cluster when the relevant MySQL rows were | updated. | paulddraper wrote: | That's so cool. ___________________________________________________________________ (page generated 2023-04-01 23:01 UTC)