Jeff Rizzo <[email protected]> writes: > The host has a NetBSD 10.1 kernel and a 9.0 userland that's in the > process of being updated to 10.1 - unpacking `base.tar.xz` is where I > first noticed problems. Here's a quick summary of the problem in > example form: > > ansible:riz ~/sets> md5 xfont.tar.xz > MD5 (xfont.tar.xz) = f044efd355a3a8fbee8988200aa526d5 > ansible:riz ~/sets> md5 xfont.tar.xz > MD5 (xfont.tar.xz) = 07e28b41bb982b2b0a1e0b731599a246 > ansible:riz ~/sets> md5 xfont.tar.xz > MD5 (xfont.tar.xz) = b0d14f58745706db4dcf4a30d5b75175 > ansible:riz ~/sets>
Are you saying that in a domU you are running md5, and the file is in UFS on an xbd, and the xbd is backed by a zvol in the dom0? I would suggest that you boot the domU single user and repeat this, and also md5 or sha1 the zvol. I have 5 domUs of various netbsd version/arch where the xbd or wd (soem HVM, some PV, some PVH) is backed by a zvol. My dom0 is netbsd-10 and the zpool has one disk only. I have not seen any flakiness like this. (zfs does lock up when writing mmmaped files under memory pressure, but that's different) I wonder if you have troubled memory and suggest running memory diagnostics on the box.
