Re: [Nbd] bcache on NBD block devices
- To: Matthew Patton <pattonme@...34...>
- Cc: nbd-general@lists.sourceforge.net
- Subject: Re: [Nbd] bcache on NBD block devices
- From: Juan Antonio Martinez <jantonio@...1345...>
- Date: Thu, 27 Feb 2014 09:48:30 +0100
- Message-id: <1393490910.27958.6.camel@...1346...>
- In-reply-to: <egtxw9pej0indsiexwg8l1ri.1393424201742@...1518...>
- References: <egtxw9pej0indsiexwg8l1ri.1393424201742@...1518...>
El mié, 26-02-2014 a las 09:16 -0500, Matthew Patton escribió:
> Shouldn't the clients be using the bcache device for their io? So who
> cares what mode the nbd device is in just as long as it's writable?
> Then just make the bcache device read only.
Sure: the clients mounts the inner squashfs readonly, but in order to
bcache get working, the nbd-server must export read-write, or at least
with copy-on-write flag, to let the bcache layer on clients work
throught nbd device
Anyway: problem solved. Performance is not noticeabilly increased, due
to obsolete hardware in clients, but network load improvement is really
good
Juan Antonio
Reply to: