From mboxrd@z Thu Jan 1 00:00:00 1970 From: Xose Vazquez Perez Subject: Re: [PATCH] multipathd: update defaults Date: Sun, 5 Jun 2016 20:24:17 +0200 Message-ID: Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit Return-path: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Sender: dm-devel-bounces@redhat.com Errors-To: dm-devel-bounces@redhat.com To: Hannes Reinecke , Benjamin Marzinski , Sebastian Herbszt , Christophe Varoqui , device-mapper development Cc: matthias.rudolph@hds.com, sramani@pillardata.com, jim.lester@compellent.com, yacine@alyseo.com, g.shyamsundar@yahoo.co.in, sean.stewart@netapp.com, davidw@netapp.com, brking@linux.vnet.ibm.com, bartb@pivot3.com, egoggin@emc.com List-Id: dm-devel.ids Hannes Reinecke wrote: > For years I've been telling our customers to use the 'tur' checker > as the current default 'directio' will cause spurious path failures > under high load. If directio is unreliable, it must be _fixed_ or *removed* . And this text in multipath.conf.5 needs to be replaced: "readsector0 (Deprecated) Read the first sector of the device. This checker is being deprecated, please use directio instead" There are 11 devices with path_checker=directio as default option, what to do with them ? > And for several versions (years, even) the linux kernel has the ability > to correctly detect the device handler, so we should be making > 'retain_hw_handler' the default. > And if we do this we also want to enable 'detect_prio', to ensure > we pick the correct prioritizer. I do really like both features, it makes the multipath setup almost a completely unattended task. And neat for live_cd, installers of distributions, rescue environments and anywhere with a ro /etc. There are a handful of devices with prio=alua as default option, and they should be revoked to their "original" configuration. > So this patch updates the default setting to always enable > 'retain_hw_handler' and 'detect_prio' and switch to 'tur' > as the default path checker. The patch should be split in two: directio->tur and autodetect_alua. CCed: All controllers maintainers, let's see what they think. > Signed-off-by: Hannes Reinecke > --- > libmultipath/checkers.h | 2 +- > libmultipath/defaults.h | 4 ++-- > multipath/multipath.conf.5 | 4 ++-- > 3 files changed, 5 insertions(+), 5 deletions(-) > > diff --git a/libmultipath/checkers.h b/libmultipath/checkers.h > index a935b3f..374c826 100644 > --- a/libmultipath/checkers.h > +++ b/libmultipath/checkers.h > @@ -85,7 +85,7 @@ enum path_check_state { > #define READSECTOR0 "readsector0" > #define CCISS_TUR "cciss_tur" > > -#define DEFAULT_CHECKER DIRECTIO > +#define DEFAULT_CHECKER TUR > > #define ASYNC_TIMEOUT_SEC 30 > > diff --git a/libmultipath/defaults.h b/libmultipath/defaults.h > index 96f5a2c..4948271 100644 > --- a/libmultipath/defaults.h > +++ b/libmultipath/defaults.h > @@ -16,8 +16,8 @@ > #define DEFAULT_FIND_MULTIPATHS 0 > #define DEFAULT_FAST_IO_FAIL 5 > #define DEFAULT_DEV_LOSS_TMO 600 > -#define DEFAULT_RETAIN_HWHANDLER RETAIN_HWHANDLER_OFF > -#define DEFAULT_DETECT_PRIO DETECT_PRIO_OFF > +#define DEFAULT_RETAIN_HWHANDLER RETAIN_HWHANDLER_ON > +#define DEFAULT_DETECT_PRIO DETECT_PRIO_ON > #define DEFAULT_DEFERRED_REMOVE DEFERRED_REMOVE_OFF > #define DEFAULT_DELAY_CHECKS DELAY_CHECKS_OFF > #define DEFAULT_UEVENT_STACKSIZE 256 > diff --git a/multipath/multipath.conf.5 b/multipath/multipath.conf.5 > index aaaa01b..20a2304 100644 > --- a/multipath/multipath.conf.5 > +++ b/multipath/multipath.conf.5 > @@ -295,7 +295,7 @@ Check the path state for LSI/Engenio/NetApp E-Series RDAC storage controller. > .B directio > Read the first sector with direct I/O. > .TP > -Default value is \fIdirectio\fR. > +Default value is \fItur\fR. > .RE > .TP > .B failback > @@ -463,7 +463,7 @@ are supported, or the > prioritizer if not. If set to > .I no > , the prioritizer will be selected as usual. Default is > -.I no > +.I yes > .TP > .B force_sync > If set to > -- > 2.6.6