guix-patches
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[bug#28635] [bug#28198] [PATCH 1/4] vm: Add disk-image-size to <virtual-


From: Christopher Baines
Subject: [bug#28635] [bug#28198] [PATCH 1/4] vm: Add disk-image-size to <virtual-machine>.
Date: Fri, 29 Sep 2017 07:48:21 +0100

On Tue, 26 Sep 2017 09:14:28 +0200
address@hidden (Ludovic Courtès) wrote:

> Hi Chris!
> 
> Christopher Baines <address@hidden> skribis:
> 
> > On Thu, 31 Aug 2017 14:29:56 +0200
> > address@hidden (Ludovic Courtès) wrote:
> >  
> >> Hi Chris,  
> >
> > Thanks for reviewing these patches Ludo, unfortunately its taken me
> > nearly a month to get around to replying.
> >
> > Thankfully, I've made some progress in that time.
> >  
> >> Christopher Baines <address@hidden> skribis:
> >>   
> >> > * gnu/system/vm.scm (<virtual-machine>): Add
> >> >   disk-image-size.
> >> >   (port-forwardings->qemu-options): Use disk-image-size from
> >> >   <virtual-machine>.    
> >> 
> >> In which case is it useful?  Perhaps if you want to create lots of
> >> data on the root file system in the MongoDB test?
> >> 
> >> Currently <virtual-machine> builds a shared-store VM (like ‘guix
> >> system vm’) in which the root file system has a fixed size that’s
> >> usually good enough.  
> >
> > The best answer I have at the moment is that I think MongoDB
> > creates a rather large file, even if it has no data to store in it.
> > I'll do some more investigation to confirm this though.  
> 
> OK, that’s a good reason anyway.  :-)
> 
> >> > --- a/gnu/system/vm.scm
> >> > +++ b/gnu/system/vm.scm
> >> > @@ -653,6 +653,8 @@ it is mostly useful when FULL-BOOT?  is
> >> > true." (default #f))
> >> >    (memory-size      virtual-machine-memory-size   ;integer (MiB)
> >> >                      (default 256))
> >> > +  (disk-image-size  virtual-machine-disk-image-size   ;integer
> >> > (bytes)
> >> > +                    (default (* 70 (expt 2 20))))    
> >> 
> >> I think we can use 'guess here as the default value (and we should
> >> do the same in places where #:disk-image-size has an arbitrary
> >> default.)  
> >
> > I've been looking at this in the last few days. Making 'guess the
> > default value here does mean something, and I think it's a good
> > idea.
> >
> > As far as I can tell, for this specific test, these are the 3
> > functions that are called on the way to using the actual size, and
> > the defaults they have:
> >
> > system-qemu-image/shared-store-script default:
> >   (* (if full-boot? 500 70)
> >      (expt 2 20)))
> >
> > system-qemu-image/shared-store default: 
> >   (* (if full-boot? 500 30) 
> >      (expt 2 20)))
> >
> > qemu-image default:
> >   'guess
> >
> > From my tests, if the default in the <virtual-machine> is set to
> > 'guess, then for the mongodb test the guess seems to be 0 MiB at the
> > moment, which doesn't work. I've started looking at setting a
> > sensible default in qemu-image, so that the root filesystem size
> > isn't 0 MiB.  
> 
> Indeed.  Maybe we can go with your patch as-is and investigate the
> problem with 'guess separately.  Thoughts?

That's fine with me. I've also now created a separate bug with this
patch, and another to add a default value [1].

I tested it with the memcached system test, and the disk image size was
the same plus the test still passed. I haven't checked the other tests
yet.

1: https://debbugs.gnu.org/cgi/bugreport.cgi?bug=28635

Attachment: pgpfO6fGyiZ0B.pgp
Description: OpenPGP digital signature


reply via email to

[Prev in Thread] Current Thread [Next in Thread]