From mboxrd@z Thu Jan 1 00:00:00 1970 From: Ian Campbell Subject: Re: [PATCH v2] xentop: add support for qdisks Date: Fri, 20 Mar 2015 09:57:02 +0000 Message-ID: <1426845422.21742.114.camel@citrix.com> References: <55002C5B02000091000F1DAD@prv-mh.provo.novell.com> <1426695146.14291.60.camel@citrix.com> <20150319180951.GJ1345@perard.uk.xensource.com> <550ACD3102000091000F2C93@prv-mh.provo.novell.com> Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: <550ACD3102000091000F2C93@prv-mh.provo.novell.com> List-Unsubscribe: , List-Post: List-Help: List-Subscribe: , Sender: xen-devel-bounces@lists.xen.org Errors-To: xen-devel-bounces@lists.xen.org To: Charles Arnold Cc: Anthony PERARD , xen-devel List-Id: xen-devel@lists.xenproject.org On Thu, 2015-03-19 at 13:20 -0600, Charles Arnold wrote: > >>> On 3/19/2015 at 12:09 PM, Anthony PERARD wrote: > > On Wed, Mar 18, 2015 at 04:12:26PM +0000, Ian Campbell wrote: > >> My second concern here is with the use of /var/run/xen/qmp-libxl-%i from > >> outside of libxl. I can't remember if qemu is safe against multiple > >> users of the socket. ISTR asking Anthony this before, but I don't recall > >> the answer, sorry :-( > > > > Last time I checked, only one client at a time can connect to the socket. > > If a second user want to connect to the socket, it will be blocked until > > the first one disconnect. > > This seems correct based on some of my testing. In rare cases (perhaps not so > rare for VMs under heavy I/O load), reading the socket to get the stats times out > and so xentop will report '0' for the read/write stats until the next read attempt > that succeeds. > > It looks as if we do need a second socket to qemu. I will include that in the next > patch version. Yep, it sounds like it is a requirement, thanks. Ian.