Streaming data out

With Postgres, there exists the ability to read the WAL (transaction log) byte code and transform it into JSON. Thus having say a GoLang process read it and execute events. Thereby having a real-time process.

Does a transaction log / replication log / something that can be readable exist?

Thanks

Hi Paul!

We have a transaction log, but if I am not wrong, it’s not readable by other third party tools at the moment. It’s only for MemSQL internally to use. But providing a transaction log that’s compliant with other tools is something many customers have asked for, so we are definitely aware of the need.

Thanks Jacky.

Many thanks for following up. Totally understood where MmSQL is at.

That said, if at a minimum a transaction log from the master to a single file in /var/lib/[master id]/data/logs where its readable by the command line would be awesome. Also having an option to either have the transaction log grow, or truncate to every Xkb or Xmb size.

I’m sure you would hit a certain requirement % with the release of this functionality.

You can close this question.

Memsql 7.0 can select into Kafka. Even though it’s not the change data capture feature lots of customers are requesting you can use it to stream data out from stored procedures or pipelines.

@nikita thanks for responding.

Whilst I’m sure others would welcome that. Personally I find this next to useless for my use case and please note that some customers do not use Kafka because it creates a lot of overhead, for many use cases its utter overkill and is akin to a black box.

As elon musk says, if it’s tight its right, if its long its wrong. When I look at infrastructure the simpliest solution is always the right one. Using over engineered software from the apache foundation will always get you into trouble!

Many thanks for replying, especially on a Saturday. Have a good one :slight_smile:

We will build CDC soon. We also are looking at continuous queries… Thanks for candid feedback :slight_smile:!

CDC would be an utter game changer. With continous queries, you guys are trying to replace Redis and maybe rethinkDB?

Definitely a massive consumer of realtime events into and out of a db. So definitely looking with much interest!

When you take CDC out where do you envision the stream to go to?

At a minimum I envision the stream to go to a background process.

But if there are options available. There should be a way of pushing to a message queue like nats.io. I’m sure others would want zeromq, rabbitmq, kafka, etc. That would then allow for multiple background processes waiting. Which would also solve redundancy issues.

So in my case it would be a golang executable reading the stream. Which contains logic that depending on what happens, it would then create another job to execute a process or send something to the FE.

Having this ability makes the process fully reactive. Rather than having another golang executable hitting a table every second and checking to see if conditions are met. Which is something I’m actually doing now for my job queue ://

The more I think about it. Having a stream is the holy grail for me lol. There’s so much that I could do with it!

Even something like this lol: https://rxdb.info/

Hopefully all this makes sense.

Btw thanks for the dialog. Much appreciated!

Last question. Is there a timeframe for CDC? Something expected due to v7?

Thanks