From mboxrd@z Thu Jan 1 00:00:00 1970 From: Ian Campbell Subject: Re: [PATCH] libxl: Increase device model startup timeout to 1min. Date: Mon, 29 Jun 2015 15:51:57 +0100 Message-ID: <1435589517.32500.342.camel@citrix.com> References: <1435319840-25020-1-git-send-email-anthony.perard@citrix.com> <1435336867.32500.209.camel@citrix.com> <20150629142317.GB1891@perard.uk.xensource.com> Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: <20150629142317.GB1891@perard.uk.xensource.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: Anthony PERARD Cc: Wei Liu , Stefano Stabellini , Ian Jackson , xen-devel@lists.xen.org List-Id: xen-devel@lists.xenproject.org On Mon, 2015-06-29 at 15:23 +0100, Anthony PERARD wrote: > On Fri, Jun 26, 2015 at 05:41:07PM +0100, Ian Campbell wrote: > > On Fri, 2015-06-26 at 12:57 +0100, Anthony PERARD wrote: > > > On a busy host, QEMU may take more than 10s to start. > > > > Please show your working here so that in 2 years when we want to know > > why this value was chosen we don't have to go scrobbling around in the > > ML archives looking for the data you gathered. > > OK. What about: I'm afraid none of that really explains why QEMU taking >10s is reasonable/expected under the circumstances or why 60s is a good choice now. Nor does it really answer Ian's question in <21901.33163.547929.321814@mariner.uk.xensource.com> I think. Ian. > > While running the whole stack of OpenStack (install via devstack on Ubuntu) > on a single machine, sometime the device model fail to start on time. When > this happen, a strace of QEMU shows that it still loading its libraries. > > You can find such strace on: > http://lists.xenproject.org/archives/html/xen-devel/2015-06/msg03549.html > > What shows the strace is that between a mmap() syscall and the next > syscall, 0.5s might have pass. > > Normaly, on this same machine it takes 0.32 second on average for QEMU to > start, while running Tempest (which is the OpenStack test suite). Those > QEMU are runned as backend for a PV guest. > > > > > > > Signed-off-by: Anthony PERARD > > > --- > > > tools/libxl/libxl_internal.h | 2 +- > > > 1 file changed, 1 insertion(+), 1 deletion(-) > > > > > > diff --git a/tools/libxl/libxl_internal.h b/tools/libxl/libxl_internal.h > > > index e96d6b5..3a604f7 100644 > > > --- a/tools/libxl/libxl_internal.h > > > +++ b/tools/libxl/libxl_internal.h > > > @@ -85,7 +85,7 @@ > > > #define LIBXL_INIT_TIMEOUT 10 > > > #define LIBXL_DESTROY_TIMEOUT 10 > > > #define LIBXL_HOTPLUG_TIMEOUT 10 > > > -#define LIBXL_DEVICE_MODEL_START_TIMEOUT 10 > > > +#define LIBXL_DEVICE_MODEL_START_TIMEOUT 60 > > > #define LIBXL_STUBDOM_START_TIMEOUT 30 > > > #define LIBXL_QEMU_BODGE_TIMEOUT 2 > > > #define LIBXL_XENCONSOLE_LIMIT 1048576 > > > > >