Jean Boussier
Jean Boussier
@rafaelfranca, since this came up again, what do you recon we should do here? I really don't see any way to warn users :/
> definitely sucks that we're breaking folks and adding more work for you It really isn't that bad, it's mostly confusion I think.
No worries. That's much appreciated.
> What we re thinking of doing is going back to the original monkey-patching for 5.x, that same that we do for 4.x. So you wouldn't instrument raw `redis-client` usage...
Note that `redis-client` can be used standalone, it's not exclusively a dependency of `redis`.
I have an open PR that adds a predicate on errors to know whether they are final or not: https://github.com/redis-rb/redis-client/pull/255 I'd love to know if this would help `dd-trace-rb`
https://bugs.ruby-lang.org/issues/21268 is what should fix it
> You would need to wrap all non-idempotent commands in MULTI then? It would be purely optional, but if you wanted to be able to retry in all circumstances yes....
> Are those primarily for clusters? Not as far as I know, they are used for a whole lot of things. > the server can end up using a lot...
> how to set the random-id? If this is really random, how to guarantee the ids are not duplicated in short time? It's the responsibility of the client to use...