Clay Gerrard fa0c5f244f Additionally break container servers during fast-post probes ...

The existing probetests were already doing brainsplitting on the object
servers in order to introduce composite metdata timestamp reconciliation.

But they were not introducing failures at the container server level, and
therefore not covering functional testing of those servers replication
consistency repair.

This change adds a brain splitter for the container servers to additionally
exercise the consistency engine at that layer under the existing probetest so
that you can observe their correct behavior while exercising the probetests
[1].

1. I used a script like this to watch the object servers and container db's
repair themselves -> https://gist.github.com/clayg/a8077d9d29ff68bb0edd

Change-Id: Ic5bb03dffef7e30c58338cef969a105577b84620

2016-03-07 13:29:10 +00:00

34 KiB

Raw Blame History

View Raw