Date: Tue, 19 Apr 2011 10:15:26 +0200 From: Borja Marcos <borjam@sarenet.es> To: Artem Belevich <art@freebsd.org> Cc: freebsd-fs@freebsd.org Subject: Re: Prompt to synchronize two volumes ZFS Message-ID: <3596CB9B-7996-404B-AD34-D3A8DFD67FD4@sarenet.es> In-Reply-To: <BANLkTim-zK1cgjbaqDFQfmXWNVD4rCLEkw@mail.gmail.com> References: <4DAC7811.3090407@ukr.net> <BANLkTinsvVzaY9yiQ5QvVGK7gg7cPFMRcA@mail.gmail.com> <4DAC96EA.8080505@ukr.net> <BANLkTim-zK1cgjbaqDFQfmXWNVD4rCLEkw@mail.gmail.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On Apr 18, 2011, at 10:15 PM, Artem Belevich wrote: > They mentioned performance. mbuffer in-between receive and send makes > *a lot* of difference as long as you provide few seconds worth of > buffering at the rate your filesystems can sustain. I think the > authors of the page above just didn't use large enough buffer. You > would probably have to experiment yourself. In my case of ~3TB > transfer (mostly large files), I ended up with "mbuffer -m512M". I > also used mbuffer's built-in network transfer mechanism (see mbuffer's > -I/-O options) as at high data rates ssh became the bottleneck. Moreover, although ZFS receive seems to be robust in case a replication = stream is interrupted, I find it much more safer to move the stream = beforehand, and start the zfs receive with a complete stream. Borja.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?3596CB9B-7996-404B-AD34-D3A8DFD67FD4>