Re : Re : Re : task btrfs-transacti:651 blocked for more than 120 seconds

[Olivier Bonvalet <mailinglist@daevel.fr>]

Le vendredi 29 septembre 2017 à 10:20 +0100, Roger Pau Monné a écrit :
> On Thu, Sep 28, 2017 at 05:27:54PM +0000, Olivier Bonvalet wrote:
> > Le jeudi 28 septembre 2017 à 17:12 +0100, Roger Pau Monné a écrit :
> > > 
> > > Quite likely. With so many PV block devices attached you either
> > > have
> > > to limit the number of queues and persistent grants per-device,
> > > or
> > > expand the number of grants allowed by Xen.
> > > 
> > > Can you try to set the following in the Xen command line [0] and
> > > see
> > > if that solves your issues:
> > > 
> > > gnttab_max_frames=64
> > > 
> > > Roger.
> > > 
> > > [0] http://xenbits.xenproject.org/docs/unstable/misc/xen-command-
> > > line
> > > .html
> > > 
> > > _______________________________________________
> > > 
> > 
> > Oh, from the Novell's documentation [0] I read :
> > 
> > « General recommendation for determining the proper value for
> > "gnttab_max_frames" is to multiply by 2 the number of attached
> > disks. »
> > 
> > 
> > Since I have about 250 RBD devices, I suppose I have to try
> > directly
> > with gnttab_max_frames=512, right ?
> 
> Do you have 250 devices attached to the same guest? If so I guess 512
> might be sensible, although you said earlier that you had 26 devices
> attached, not 250.
> 
> Roger.
> 

Hi,

no VM have more than 26 devices. Except for the Dom0 which have about
300 devices to handle. Does dom0 affected by this gnttab_max_frames
choice ?

Anyway, after booting each hypervisor with gnttab_max_frames=256 (or
greater), it seems I don't reach this timeout anymore.

Thanks !

Olivier



_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
https://lists.xen.org/xen-devel