Re: Staccali-install notes

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

 



On Wed, 2006-07-26 at 15:10 -0600, Jeffrey Law wrote:

> I can't get staccali-install to work from a readonly-root; the LVM 
> programs used by stacaccli-install want to scribble /etc/lvm/.cache
> which fails (readonly root).  We should probably add that to the
> list of necessary bind mounts.

	lvm actually handles this fine, so long as the gets an EROFS when it
tries to create the file. Unfortunately, it gets an EPERM on NFS. I
spent some time trying to come up with a fix in NFS for that, but didn't
get anywhere. AFAIR, the EPERM is returned by the server.

	Perhaps LVM should just handle EPERM in the same way.

> When staccali-install wants to create an initrd, it runs mkinitrd, which
> in turn reads /etc/fstab to determine characteristics of the root
> filesystem.  More likely than not /etc/fstab is going to be totally
> inappropriate.  Instead staccali should conjure up its own fstab for
> use by mkinitrd (we know where the root filesystem will be, so this
> should be trivial).

	Yeah, this goes back to the mail I sent out a while back about "what to
do with an fstab on an image shared between diskless and cached
clients".

> We need to make sure the proper /dev/XXX files are in place so that 
> when mkinitrd scans for logical volumes, it'll find some and in 
> turn add the proper lvm bits to the initrd (staccali might handle
> this correctly, this is something I found out after making my own
> initrd when staccali's failed).

	Yeah, stacaccli has code to create the device node in the chroot.

> I needed an "import sys" in one of the staccali files.  I don't
> immediately recall which one, but it'll be easy enough to find when
> I re-install staccali.
>
> As noted in an earlier message, we leak FDs (not terribly important)
> when downloading the image from the server fails.

	Yep, feel free to fix both of these.

> staccali-install should be setting up the local scratchpad and 
> state filesystems as well, it should probably create /home with
> whatever space remains.
> 
> Integration with puppet -- we basically just need to inform the
> puppet server that a new host is under puppet management.  That
> should trigger some server-side things like creating ssh keys
> for the host and puppet certs and putting them in the server's
> database of files it manages for the client.

	Don't really have an opinion on either of these. But it sounds fair
enough.

> After using the same image created by my trivial imaging code for
> both diskless clients and for caching clients, there's some 
> minor updates to make to the imaging code.  The good news is that
> I'm able to use the same image for both usage cases :-)

	So, what changes were needed?

Cheers,
Mark.


[Index of Archives]     [Kernel Newbies]     [Share Photos]     [IDE]     [Security]     [Git]     [Netfilter]     [Bugtraq]     [Yosemite News]     [MIPS Linux]     [ARM Linux]     [Linux Security]     [Linux RAID]     [Linux ATA RAID]     [Samba]     [Linux Media]     [Device Mapper]

  Powered by Linux