kernelOfTruth aka. kOT, Gentoo user

ugh :-1: zdb -M WD30EFRX vdev 0 metaslabs 170 fragmentation 0% pool WD30EFRX fragmentation 0% space map refcount mismatch: expected 167 != actual 0 just migrated to 3.19 and latest...

related: https://bugs.freenas.org/issues/6352 https://forums.freenas.org/index.php?threads/problems-with-freenas-9-2-1-5.20879/ so it's not limited to ZFSonLinux, but also occurs on FreeBSD \ m0n0wall \ NanoBSD \ FreeNAS - based systems

luckily it finished without errors here: ``` zpool status -v WD30EFRX pool: WD30EFRX state: ONLINE scan: scrub repaired 0 in 6h14m with 0 errors on Mon Feb 16 03:40:04 2015...

zdb -mc says: Traversing all blocks to verify metadata checksums and verify nothing leaked ... ``` 1.69T completed ( 597MB/s) estimated time remaining: 4294792253hr 4294967281min 4294967292sec leaked space: vdev 0,...

@behlendorf lots of vdevs use ditto blocks on my pools, could data integrity for those files also be compromised in combination with this issue ? what features are heavy users...

an example how this manifests on newly created zvols/zpools: ``` zpool create -o ashift=12 SWAP2_1 /dev/mapper/swap2_1 -f zpool create -o ashift=12 SWAP2_2 /dev/mapper/swap2_2 -f zfs set atime=off SWAP2_1 zfs set...

ugh :( related: http://lists.freebsd.org/pipermail/freebsd-current/2011-October/028232.html http://lists.open-zfs.org/pipermail/developer/2014-June/000691.html not sure if related but referencing #2662 and #2702

@edillmann yes, I used #2129 since that improved the desktop improvement quite noticably without it (now) - there are some increased minor latencies here and there Did you try #2129...

like mentioned in the above linked issue, this system uses glibc-2.18 (glibc-2.18-r1 on ~amd64 Gentoo GNU/Linux)

installing glibc-2.19 or newer should fix the issue, it did at least for me - Portal 2 launches now