Description
The disk-failure workload in Jepsen is currently broken. The workload needs to be adapted to the reworked nemesis logic from MB-35752. However, the current logic in the nemsis has been likely broken since commit 3311437 and 863fbb3. Those commits attempted to introduce recovery following a disk failure, but this is likely not possible. Once the virtual disk has been set to hard fail IO errors, the filesystem will likely detect the errors and lock the fs. It is therefore not possible to reliably recover from this situation.
Attachments
For Gerrit Dashboard: MB-35759 | ||||||
---|---|---|---|---|---|---|
# | Subject | Branch | Project | Status | CR | V |
117293,2 | MB-35759 Fix fail-disk nemesis action | master | jepsen.couchbase | Status: MERGED | +2 | +1 |
117294,3 | MB-35759 Fix disk-failure workload | master | jepsen.couchbase | Status: MERGED | +2 | +1 |