Just found a bug in jepsen.tests.long-fork which (often) allowed it to declare histories containing long-fork anomalies as valid, instead of invalid. Jepsen 0.1.14-SNAPSHOT, available on Clojars, contains a fix.https://github.com/jepsen-io/jepsen/commit/5517595082129066cfe1166718102b8e6837f431 …
-
-
I've taken a few hours to rerun Dgraph's test suite against 1.0.13, and YugaByte DB's against 1.1.15.0-b16 with a corrected long-fork checker, and neither exhibited long fork anomalies with the standard set of of failure modes. I've noted this in errata on the respective reports.
Show this thread -
Usual disclaimer applies; this isn't a proof that long-fork *can't* happen; just that this particular checker bug doesn't seem to have caused us to miss issues. There *was* an instance of long fork in the original Dgraph report, but it was actually identified by a different test!
Show this thread
End of conversation
New conversation -
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.