RT @asatarin
"Cobra: Making Transactional Key-Value Stores Verifiably Serializable"

Given black box trace of KV transactions determine if observed behaviors are serializable at scale (10K transactions)


New report: we worked with @ScyllaDB to identify and fix cases of LWT split-brain in healthy clusters due to improper hashing and membership changes, as well as documentation improvements. Notably, Scylla no longer claims non-LWT isolation! jepsen.io/analyses/scylla-4.2-

RT @yow_conf
Can you believe it?! We're adding ONE more speaker to and it's @aphyr!

We trust databases to store our data, but should we? Learn the basics of distributed systems testing & advice for testing your own systems in his keynote - Jepsen 13.

RT @adriancolyer
"Elle: inferring isolation anomalies from experimental observations", Kingsbury & Alvaro, VLDB'20. blog.acolyer.org/2020/11/23/el

Be afraid databases, be very afraid: Jepsen just got a whole lot more powerful...

Jepsen's AWS Marketplace product (launches a whole Jepsen cluster in a few clicks) has been updated for Debian Buster, and now includes additional dependencies you'll want for running tests based on Elle. Happy testing!


Put a bunch of work into the Jepsen docker-compose setup, and I'm pleased to report it now gives you a full Debian Buster cluster with keys and dependencies out of the box--should be good for running the latest tests. Hopefully this helps!


Do you use Cassandra or another CQL-compatible database? I'd like to hear your perspective on adding things to a CQL set: docs.google.com/forms/d/e/1FAI

A bug in Jepsen: from versions 0.1.2 to 0.2.0, the counter checker docstring incorrectly claimed to handle decrements, which could cause valid histories to be reported as failures. This did not affect official Jepsen reports, but other counter tests using decrements may have been affected: groups.google.com/u/1/a/jepsen

Love this thing where Google Cloud decides that jepsen.io has been stable for a while and it really ought to do something about that, so it kills the VM and spins up a new one to replace it only *after* it's dead, resulting in ~10 minutes of spurious downtime.

It's been doing this for ~two years, COME ON Google, y'all are supposed to be experts at rollouts. Start new nodes *before* you kill existing ones!

Did an interview with Tobias Macey talking about Jepsen's design, software verification in general, and the distributed database landscape: dataengineeringpodcast.com/jep

RT @andy_pavlo
Reminder: The @jepsen_io Quarantine DB Talk featuring Kyle Kingsbury (@aphyr) is next Monday July 27 @ 4:30pm ET. Video will be live and uncut for the public over Zoom. All are welcome to join. db.cs.cmu.edu/events/db-semina

Anyway, if you have strong feelings, drop em here.

Show thread

It's also, like... Jepsen is roughly 50/50 paid vs unpaid work right now. Jepsen contract rates are high, which covers research, maintenance, and writing in between. It's hard to imagine sponsors could materially shift that balance.

Show thread

On the other hand, this presents a conflict-of-interest problem: so long as reports have a single sponsor (typically the vendor), it's easy to disclose and understand, but that's much trickier when there's a mix of a dozen ongoing sponsors.

Show thread
Show older

A single-user Mastodon instance for Jepsen announcements & discussion.