MongoDB is introducing pluggable storage engines in 2.8. I was wondering if there's been any thought given to having a TokuMX / fractal tree storage option? This would be much more similar to the way that TokuDB is implemented as a storage engine for MySQL.
We have an active project to merge the 2.6 release which you can follow here. We also plan to have a storage engine but no date has been assigned. TokuMX will still be offered.
That's awesome! I'm really excited by that. If there's a JIRA tracker made for it at some point I'd love to follow along.
Current status can be found at https://github.com/leifwalsh/
Btw, our MongoDB Meetup Group (>400 people) is holding a bug hunt next week for the new storage engines. If it was far enough along that you wanted help testing the Toku storage engine we could help with that. Sounds like it might be a bit early for that, but thought I'd offer anyway in case it could help.
Thanks for the offer. It sounds like something we'd be interested in. When is it? We should do a bit more triage and polishing before putting it in your lap I think.
We're doing it at 6:30 pm PST this Thursday.
I think we'll have a build we can give you to test. It's somewhat bare-bones right now---the only option is --storageEngine=tokuft, I can't guarantee everything works correctly, and I know for sure some things are still a bit weird especially around concurrent range operations like db.coll.remove()---but it should work for some basic testing, and hopefully you all can help us track down some more issues.
https://s3.amazonaws.com/
I can provide an S3 link to a tarball, is that sufficient for you all?
Cool, sounds good. If you're able to get it up a bit earlier in the day I can try to test it before the event to make sure I know how to set it up.
Hey everyone, we've got a build for you. It is *very* hot off the presses, so don't burn yourselves. There are definitely some performance weirdnesses, for example jstests/core/removeb.js, and some other known failures, but let us know what you find and think. The only option right now is --storageEngine=tokuft.
I'll be up for a little longer (but I'm on the east coast) if you all need help getting it going.
Wow!
Sweet mother of perl! Can't wait to have a crack at it!
I'm thrilled to see all your enthusiasm, but for the record this is *super experimental* and very well may fill your house with cats if you look at it wrong. Normally we wouldn't release it in this state but figured getting Ben's group to take a peek was worth it. We'll definitely announce again when we have something of higher quality.
As much as I support stability. I would love to see nightly builds of certain features (like this one). It's sometimes good to have a peek at the future no matter how far/uncertain/unpolished/
Thanks for allowing us to test Leif!
Just a heads up to anyone else who might want to try it, you should make sure to use a driver compatible with 2.8, which at the present time is only the release candidate drivers. I was using the latest released driver, which isn't fully compatible with Mongo 2.8 and ran into some issues there that prevented me from testing further.
We ran into a handful of issues including the driver issues and didn't find any real bugs in the Toku stuff, but hopefully we can address those and give the testing another go. It also turns out that the company hosting us for the bug hunt was a Windows shop and so a lot of the folks there had Windows machines and could only test the vanilla Mongo, which wasn't something I expected. It seems like a pretty good part of the functionality is working though since we didn't find any issues in what we were able to test, which is pretty cool =)
Here's a new build Leif made, which works a lot better for us. We ran our integration test suite against it and it passes with this build now (wasn't quite passing with the last build yet).
댓글 없음:
댓글 쓰기