Chris Siebenmann

We're not using Lustre; it's all straight NFS (originally v3, now mostly v4). We routinely ask ZFS for all user quota information, so I think raw quota information must be...

If I'm reading the code correctly, using zdb to search for this requires using `zdb -Gddd ....` and then looking at the end of the output for the output from...

Here is what I think I see about what is happening in the code, to try to keep track of this. * the panic happens when a dnode's bonus data...

Unfortunately we cannot reproduce this issue under any sort of controlled conditions (and we've tried). This panic happens only rarely and is extremely unpredictable. We don't know the trigger conditions,...

These servers are our production ZFS fileservers, so we're not in a position to run current git master on them in order to test things.

All of the pools involved in the crashes are new, only the filesystems themselves are old. Authentic versions of test filesystems probably can't be created now except by pre-2012 or...