core dumps onto ZFS

[Mikhail T.]

Hello!

Last night I was trying to get KDE5 to start up on my new machine, and a 
couple of KDE's processes kept crashing, dumping cores like the one below:

    -rw-------  1 mi    wheel  45780992 Jul 23 22:28 ksplashqml.core

After, maybe, 10 such rounds -- each generating two core-dump -- ZFS 
hung... The machine was otherwise responsive, but any attempts to access 
the ZFS filesystems would hang as NFS would, when the remote server 
stops responding...

Pressing Ctrl-T would show the process in the state named "zfs". 
According to "systat -vm", all four disks involved in the raidz1 were 
writing in excess of 100MB/s, so I let it be for a few minutes, but 
nothing improved -- and the writes continued...

I pressed Ctrl-Alt-Del, which initiated a shutdown, but the shutdown 
hung as well ("some processes would not die") and I had to do a power 
cycle...

The sole zpool consists of 4 3TB drives and a 16GB log (on an SSD) thus:

             NAME        STATE     READ WRITE CKSUM
             aldan       ONLINE       0     0     0
               raidz1-0  ONLINE       0     0     0
                 da0     ONLINE       0     0     0
                 ada1    ONLINE       0     0     0
                 da2     ONLINE       0     0     0
                 da1     ONLINE       0     0     0
             logs
               ada0e     ONLINE       0     0     0

It reports no data-errors after reboot. There are multiple filesystems 
on it, among them /home. The box is running a very recent 
FreeBSD-11/amd64 (r336626). It has 4 Xeon cores and 128GB of RAM. The 
pool was created under FreeBSD-10 -- after this incident I upgraded it.

What happened? Thanks! Yours,

    -mi