For what it's worth, even after merging this change to ZoL we are still seeing rare `ztest` failures where reconstruction is either impossible or possibly exceeds the allowed combinations. I don't see an issue with the reconstruction logic, it _may_ still be possible that `ztest` can damage a pool beyond repair. For the curious, here are links to the [ztest log](http://build.zfsonlinux.org/builders/CentOS%206%20x86_64%20%28TEST%29/builds/4358/steps/shell_6/logs/1.ztest.out) and an [ztest pool](http://build.zfsonlinux.org/artifacts/pull/7939/zfs-0.8.0-rc1-39-g6458894/CentOS-6-x86_64/ztest/ztest-20181008T134630.tar.xz).
-- You are receiving this because you are subscribed to this thread. Reply to this email directly or view it on GitHub: https://github.com/openzfs/openzfs/pull/625#issuecomment-428014366 ------------------------------------------ openzfs: openzfs-developer Permalink: https://openzfs.topicbox.com/groups/developer/Tf29308eb1ca3d052-Mba42fc08126899c90779d6c1 Delivery options: https://openzfs.topicbox.com/groups/developer/subscription