Re:How Discord Stores Billions of Messages Using Cassandra
by Vishnevskiy on Thursday January 19, 2017 @06:09PM
(#13440328)
The problems that Jepsen found were centered around the "transactions" feature that Cassandra added. We don't use these and don't need them since we don't need 100% consistency and prefer availability (for example we read at quorum to trigger read repair, but downgrade to single node reads if we need to).
Also ScyllaDB is a new product and it would be crazy to start off with it. We plan to run a long-term double write experiment before we are comfortable with using it as a primary data store.
Re:How Discord Stores Billions of Messages Using Cassandra
by smharris65 on Thursday January 19, 2017 @08:51PM
(#13441168)
The Jepsen tests were not completely centered around transactions. It also had to do with data loss when replicas go down and pure "last-write-wins" approach. For those wanting more info around this the original post is here:
Re:How Discord Stores Billions of Messages Using Cassandra
by Vishnevskiy on Thursday January 19, 2017 @08:52PM
(#13441173)
Last-write-wins is a semantic we are okay with for this data, and dealing with one of our conflicts is outlined the article.
Re:How Discord Stores Billions of Messages Using Cassandra
by smharris65 on Thursday January 19, 2017 @09:11PM
(#13441241)
I enjoyed your article and I do appreciate your transparency!
Re:How Discord Stores Billions of Messages Using Cassandra
by jjirsa on Thursday January 19, 2017 @10:51PM
(#13441615)
I find it fascinating that people still think Cassandra is some risky new tech - been running it in production since 2010, and the fact that people are still worried about it makes me snicker a bit.
Re:How Discord Stores Billions of Messages Using Cassandra
by Thaxll on Thursday January 19, 2017 @06:31PM
(#13440446)
Not everyone need strong consistency, that's why those Jepsen reports are not relevant for all cases.
Re:How Discord Stores Billions of Messages Using Cassandra
by di4na on Friday January 20, 2017 @02:25AM
(#13442224)
The whole ideas behind Jepsen report is not that people need Strong Consistency. It is that products should tell you precisely what they guarantee or not.
This is why Aphyr always test claims.
Re:How Discord Stores Billions of Messages Using Cassandra
by manigandham on Friday January 20, 2017 @11:33PM
(#13448811)
We're running Scylla in production. Much easier to setup, no tuning necessary, great performance, no issues. It has a strong team behind it.
Missing some features that Cassandra has that should be fixed by v2. Read their site/blog for a good overview of the tech and progress.
Re:How Discord Stores Billions of Messages Using Cassandra
by cnlwsu on Friday January 20, 2017 @09:43AM
(#13444081)
Cassandra now runs jepsen regularly see. If you dont believe they are fixed you can check yourself: https://www.youtube.com/watch?v=OnG1FCr5WTI
On the topic of looking at Scylla next, I wonder why did the team not just start out with it to begin with. Also, are they people with experience running both. How is the performance? And what is the state of reliability?