zfs sends "hang" recipient?

Rainer Duffner rainer at ultra-secure.de
Sat Apr 19 09:27:12 UTC 2014


I’ve got a FreeBSD 9.1 server that does zfs sends (via zxfer from ports) to a FreeBSD 10.0 server. All AMD64, the pool is some 6T, 50% full. Sender has 144GB RAM, recipient has 192GB.

The snapshots are done with one of the tools in the ports (can’t remember the name right now, I’ve basically tried them all).
During the send process (with takes 5 to 15 minutes), the receiving host is blocking all commands involving filesystems.
Stuff like df, or zpool list hangs until the receive has completed.

Is this a known problem?

Would it help upgrading the 9.1 server to 10.0, too?
In this setup, I’m really conservative regarding upgrades, as I have very little downtime.

I’ve got a MySQL database on the receiving server (not on the same filesystem, of course, but on the same pool) that functions normally.

I could ignore this, but our statistics-gathering tool relies on stuff like the above and is completely b0rked by the hangs (which I could ignore again, if it wasn’t for the customer who wanted these statistics, too…)


More information about the freebsd-fs mailing list