ILooks like it is! Thank you 
IAny insights on what happened, that you are willing to share? 
MNot yet. We restored and are still working on the root cause.
Sd1 supports how many data type? when I create the table there is even no varchar, only text.
CIt should be all sqlite data types which are NULL, INTEGER, READ, TEXT and BLOB
Sso d1 is sqlite?
CYes
Sthanks
SI have a large complex MySQL application (like erp), is it smart to migrate to D1?
CHow are you connecting to the MySQL DB? There isn't a connection string that you can use to connect to a D1 DB externally, done through a worker or HTTP
SI originally plan to use PHP+MySQL. If I migrate MySQL to D1 then I have to migrate PHP to Worker.
Stested. D1 performance is not good.
HAre you using the API? That is usually much slower
Syes, I am using Page Functions to insert data into D1
HThere, the performance also depends on how close/far away the DB is
SYou will never get faster than on-prem perf with a remote DB
SIf DB perf is a requirement, you might want to look into on-prem. As far as I'm aware, D1 offers easy scalability (due to CF having 200 trillion servers, real number btw)
Rhi ummm, I couldn't find current D1 limitation article? could anyone mind to share it
Roh, there is pinned message
RI see then, let me stay with MongoDB Atlas
IIs there a way to see per table sizes (and even index, possibly) in D1?
II know
dbstat
HI would note that D1 can(once replication goes live) theoretically be faster than on-prem if you still use Workers. As in, you would have to have/managed many DBs to be able to match the scale/locations of D1
SAh yes, I was speaking from the perspective of him using php remote to D1
SWe fully transitioned from mongodb to d1 yesterday and it's plenty fast! The python script that runs on our server updates the DB while our pages functions act as the client which return the results on our api. Considering that many requests happen every minute which receive the same output how could we cache the response for example one minute in order to not burn row reads?
CUse the cache API to cache responses. The function will still run but will reduce usage of D1
JHey all, I'm new here. I'm noticing really slow performance using D1 (cold and hot runs) in other countries like Germany and Australia. I'm getting ~1200ms on cold starts and ~700ms on hot. Is this normal for the beta?
IWhere is the D1 located? Or better yet, where have you created it from?
ITrue, I assumed the current beta database version, but yeah…
JTo clarify, my queries were just for reads.
Most recent version, I think? Just created the D1 on Oct 1st and testing with a deployment made yesterday. I think it was in us-east.. but not sure where to check the region.
Most recent version, I think? Just created the D1 on Oct 1st and testing with a deployment made yesterday. I think it was in us-east.. but not sure where to check the region.
IIt is the closest location, in which D1 is supported, from the location you hit when you create it. So if you are in us-east, it would have been created in and around there.
ISo a 700ms latency, with the ocean crossing might make sense, even though it seems a bit high. How did you fetch it? Via a Worker?
IYou should see actual database execution times in the query response json (under meta)
JIt's via a worker. I'll have to check the meta. Should I expect their to be any regional or edge caching with D1 in the beta? In latency testing in US-Mich, Germany, NL, and Australia, I haven't seen faster latency than a singe region RDS micro instance.
IThere will be read-replicas, but an ETA is currently unknown. You can do your own caching in the worker via the cache api
JAh, got it that makes sense for now. I'd make a note in the d1 docs to mention this as a workaround, as it wasn't clear to me if I was going to roll out an app.
IYeah, but it works and it won’t be needed in the long run (even though it’s reduce costs, so it makes sense all the time, tbh)
DHey Matt, I'll bump this message with another "use case", or more correctly, reason to implement - ORMs and libraries. Lots rely on transactions being a thing, and db.batch doesn't fully replace, as you can't split them and execute business logic inbetween queries.
Do you have sql transactions on the roadmap somewhere?
Do you have sql transactions on the roadmap somewhere?
Panyone ran into issues where migrations list works when using --local but not when trying to apply to prod? i just get empty output when running against prod (no errors). trying to migrate a new database for first time
POh, D1 isn’t actually on the edge just yet then? I had assumed the workers would be querying from the closest D1 instances, making for very fast reads. Isn’t that the main value prop of d1?
PI’m not seeing anywhere on the website where it states that the read replicas are disabled. I was looking to launch Monday so this is terrible news for me. Anyone have info on this?
OWould be great to see some work done on wrangler validation generally around naming conventions. One of my team created a table in the dashboard with a hyphen in it.
ID1 is on the edge, it’s generated closest to where it’s generated. Read-replicas aren’t here yet,during the beta.
IIt will still work, it’s just not distributed. It’s a beta after all…
IIt does seem weird it’s not written anywhere replicas don’t exist… 
JI think people expect D1 to behave more than Turso where DB read and writes are made only to a local db edge instance, which is asynchronously later sent to the primary region. This keeps writes and reads fast for the worker as it doesn't have to wait on the primary region (outside of cold starts sync).
https://docs.turso.tech/concepts
https://docs.turso.tech/concepts
Concepts for understanding how Turso works.
POk, so you’re saying the main database, which is the single source of truth, exists in the location closest to where it was created?
PYes this is what I expected! Will this be how D1 works once read replicas are live?
PI get that it’s beta, but I wasn’t expecting the main value prop to be missing at the moment. I mean… I could have kept my Postgres database and just added caching, there’s no difference right?
PYea, total gotcha. I’m bummed
