RDS would be amazing but is probably too big in sc...
# generald
Dmitry Pavluk
03/27/2021, 5:47 AMRDS would be amazing but is probably too big in scope to just implement
f
Frank
03/27/2021, 6:58 AM
Yeah, we’ve been mostly prioritizing constructs that invoke Lambda functions. Just curious, are you guys using native functions or stored procedures in RDS to trigger Lambda function?
d
Dmitry Pavluk
03/27/2021, 7:14 AMwe're using
aws_lambdaDmitry Pavluk
03/27/2021, 7:14 AMit's barebones to say the least haha, but still better than polling DB for changes
Dmitry Pavluk
03/27/2021, 7:18 AMI'm going to hook into the WAL eventually, which will serve as the origin of a CDC pipeline terminating in a data lake
Dmitry Pavluk
03/27/2021, 7:18 AMStreaming the DB logs to multiple downstream consumers should in theory replace the
aws_lambdaDmitry Pavluk
03/27/2021, 7:19 AMDmitry Pavluk
03/27/2021, 7:21 AMAurora already leverages the WAL to enable lag-free replication. Any commit to the log is transferred simultaneously from the WAL to disk and to the Logical Decoding Interface which in turns pipes it over to the replicas
Dmitry Pavluk
03/27/2021, 7:29 AMThe nondescript "experimental data store" is actually the reason I rearchitected half the stack, which in turn led to SST!
t
thdxr
03/27/2021, 2:13 PMDo you need an sst construct for RDS? Can't you just spin one up in a public vpc for dev?
d
Dmitry Pavluk
03/27/2021, 7:05 PMDon't need one - we've been operating on an Aurora cluster spun up via CDK construct. It's a nice to have
t
thdxr
03/27/2021, 7:29 PMI guess a better question is what be what would an SST aurora cluster do over the base one?
3 Views