It's been a long time since I've remembered to write what used to be the weekly cassandra wrapup, but I wanted to try to summarize some of the recent events (especially NGCC) for those that couldn't attend / didn't know it was happening / don't have time to follow every conversation in the community.
First: - It's been a long time since we've cut releases. There are active votes for 2.1 and 2.2, and there are upcoming votes for 3.0.15 (~66 fixes) and 3.11.1 (~72 fixes). Pretty happy (personally) with the state of the 3.0 and 3.11 branches following the last few months of fixes - if you're eagerly waiting for 3.0.15 or 3.11.1, I suspect the wait will be worth it. NGCC: - A few dozen developers gathered in San Antonio this past Tuesday to talk about future goals of / ideas for the project. Huge thanks go out to Eric and Gary (long time members of the PMC) for organizing, and Instaclustr for sponsoring everything to make this happen smoothly. People are working on getting notes up at https://github.com/ngcc/ngcc2017 - if you gave a presentation and are able to upload it, please remember to send a pull request. Here's a quick summary (from my flawed memory) of the longer talks - the lightning talks may show up in a future email when I have more time: - Ben spent some time talking about the state of the community, quantifying with graphs what many of us already know (some of our longest committers have ramped down in contributions, but we still have a pretty active committer base, and a number of new contributors ramping up). - I spent some time talking about virtual tables, and why I think CASSANDRA-7622 got buried in a bike shed full of ponies - Dikang talked more about pluggable storage engine, quantifying some gains they've seen from putting RocksDB beneath cassandra, about some of the ways they've modified cassandra to make modeling clustering work on Rocks, and about how they think streaming will work. - Ariel and Sushma both talked about decoupling durability / availability and improving latency/availability tradeoffs with a concept of transient replicas and persistent hints, where writes go to hint-only members of the cluster and are merged on reads. I'm not going to try to explain this in an email, https://issues.apache.org/jira/browse/CASSANDRA-13442 is worth reading. Maybe we'll get some notes online so people who missed it can catch up on the details and differences in the two proposals. - Avi talked about some challenges in the way Scylla implemented TPC. - Jay and Simon talked about how they're using CDC for ingestion, and proposed some changes to make things lower latency and easier to implement for CDC consumers in the future (and their talk is online: https://github.com/ngcc/ngcc2017/blob/master/CassandraDataIngestion.pdf ) One of the take-aways from NGCC (coming up first in my talk, and at least once after that) was the belief that we need to start marking experimental features as experimental. There are some users who KNOW what's safe to use, but there are many who don't realize that not all features are equally well tested. This conversation will likely spawn a number of emails like the one from Blake today, where he proposed marking MV's as experimental ( https://lists.apache.org/thread.html/d81a61da48e1b872d7599df4edfa8e244d34cbd591a18539f724796f@%3Cdev.cassandra.apache.org%3E ) . Somewhat similarly (but also very different), Sylvain has suggested deprecating the user-configurable tuning options for async read repair - https://issues.apache.org/jira/browse/CASSANDRA-13910 . If you feel strongly this is the wrong thing to do, you should probably make your voice heard sooner than later (I suspect we'll also see this on the user+dev lists, not buried in this email, so more people are aware of this change before it actually happens). As always: there's no shortage of patch-available tickets that could use a reviewer ( https://issues.apache.org/jira/browse/CASSANDRA-13901?jql=project%20%3D%20cassandra%20and%20status%20%3D%20%22Patch%20Available%22%20and%20reviewer%20is%20empty ) , and plenty of low hanging fruit tickets ( https://issues.apache.org/jira/browse/CASSANDRA-13917?jql=project%20%3D%20cassandra%20and%20labels%20%3D%20lhf%20and%20status%20%3D%20open ) . If you have a patch needing review, and the reviewer hasn't gotten to it, it never hurts to remind them (or email the JIRA to the dev@ list for extra attention). Unless I'm marked as the reviewer, in that case I haven't forgotten, I just haven't had time. Let's see if I remember to do this again next week. - Jeff
