I don't see anything obviously wrong with that

if you're just deleting a single relationship, you can also use the

Write

method with a

DELETE

operation

i'm going to try to repro, which datastore are you using?

Memory, I pulled the docker command from the github. Let me get the image hash for you

i'll just use a local process, if I can repro then we're done, if not we'll have to keep digging. There's a non-zero chance there's something weird about your wrappers.

Image hash:

730ff726f044

Run Command:

docker run --network=host quay.io/authzed/spicedb serve --grpc-preshared-key "test-auth" --grpc-no-tls --http-no-tls --metrics-addr ":9091"

No doubt about that. That's mostly why I came here. I'm literally less that 36 hours into this so I wanted to see if I was just completely misunderstanding or if there were more idiomatic ways to do things

Okay so I just restarted the docker and it worked. Which brings up a new question, are relations unique or can there be multiple of the same relation if

CREATE

is called many times? I had previously restarted the docker and it didn't work but I have also run the full test suite which would have had multiple

CREATE

calls on the same relation. I had assumed they were unique. Is that correct?

relationships are unique

Odd. So this is clearly something on my end.

As I play around with this I'll get a better sense of what I screwed up. Is this the best place to post that info so other newbies can find it?

actually it looks like memdb is not enforcing relationship uniqueness

I can repro if I write the same relationship twice

Interesting. I'll switch to postgres for testing then.

i'll try to get a fix out today

Thanks for all y'alls work on this. The power of this system is crazy for how simple it is!

What is the max size a zookie can be?

@User https://github.com/authzed/spicedb/pull/275

Thank you!

so the other datastore implementations have unique indices that prevent duplicate relationships from being created, this is something that we have to do manually for memdb

Do write operations (or delete) provide a fresh zookie on each execution since they are fully consistent by default?

yes

in postgres we're doing MVCC manually, and the write/delete will provide the new transaction ID that it synthesized

Odd, I'm getting the same one back as the prior check op. I'll switch to postgres now and see if the behavior aligns

in cockroachdb you will get back an encoded cockroach hybrid logical clock

BTW I saw there was a thought about supporting TiDB. Is that a serious consideration or just a thought? I love that database

we aren't really familiar with it, but we certainly wouldn't make an attempt to stop someone implementing a datastore driver with it

what do you like about it?

It's titanium man. I've put that system through hell. Random node failures on basically a daily basis, tens of thousands of queries per second, a dozen servers both scaled out and in. It just doesn't break. Plus the tiup client is a gift form god