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

Re: Need help with netinst of wheezy on pentium 4 machine



On 06/05/2013 03:51 PM, Bob Proulx wrote:
Hi Paul,

paul e condon wrote:
I purchased some time ago a refurb Dell GX620. I have received
conflicting advice on this list about whether I should use amd64 or
i386 CDs to install Wheezy, but amd64 CDs seem to get further into
the install so what I want help with now is all about amd64 type
CDs.
As far as installation goes either i386 or amd64 should install and
complete successfully.  There shouldn't be any difference in terms of
supported architecture for either of those.  That you are seeing a
difference is troubling and probably means that the media is corrupted
for one of them.

The difference is either 32-bit or 64-bit.  There is a lot of
conflicting advice because there isn't clearly one that is always
superior to the other.  For some things 32-bit is better.  For other
things 64-bit is better.  And therefore in the end you will simply
need to pick one and move forward with it.

I will say a few words about each.  The amd64 is newer and "cooler"
and so there has been a lot of momentum moving that direction.  It has
more cpu registers and therefore the compiler can optimize code to run
faster by making use of the extra registers.  But pointers are 64-bit
each and use more memory.  Applications that have a lot of pointers to
data need a lot more memory.  In general a 64-bit system tends to use
more memory than a 32-bit system.  Therefore if you are short on
memory then a 32-bit system is more efficient.  Memory is quite
affordable these days and an 8G 64-bit amd64 system will often be
cheaper than a 4G 32-bit i386 system.  With the more memory and
performance I would always pick amd64 at that point.  If I only had a
2G machine then I would stay with a 32-bit i386 system.

For a long time closed source proprietary applications were usually
only 32-bit.  Think Adobe Flash and so forth.  So having a plain
32-bit system often made using these 3rd party closed nonfree blobs
easier.  But since so many people are using 64-bit these days most of
those problems have been resolved in one way or another and for the
most part a 64-bit amd64 system will now run anything that a 32-bit
i386 system will run.  If that is truly a concern for you then
sticking with a 32-bit system might make things simpler.  But I only
install 64-bit amd64 these days.

I set out to install using LVM and selected a single large logical
partition. The disk has a capacity of 160GB. The install failed
because, the error message said the requested architecture was not
available on the mirror that I had chosen.
What mirror did you choose?  I routinely use ftp.us.debian.org from
the installation menu.

But it did not say what architecture it thought I had chosen.
What installation image did you use?  One of the fixed architectures
or one of the multi-architectures?

And the 'mirror' that I was invoking is my local installation of
approx on a wheezy on a different local computer.
I haven't used approx for a long time.  I can't remember if it is hard
configured for upstream sites or if it is totally transparent.

I have been using apt-cacher-ng and like it.  It is transparent,
meaning that whatever upstream you are asking for is cached and
proxied.  I liked it better than approx.

Other than that if you have CD#1 (or DVD or whatever) you should be
able to install a complete minimum system only from it.  After
installation then you can switch the /etc/apt/sources.list to your
approx proxy and install additional components.

I look into /var/cache on that computer and both amd64 and i386 are
present in the approx cache. This morning I resume work and the
first thing I want to do is try again, because I do make
mistakes. But this morning I cannot repartition the disk using the
partitioning system on the CD that was burnt from a jigdo iso image
on May 31. It would not delete/remove/whatever the LVM and would not
recognize it as a usable /root.
There is a confusing behavior of the installer and I think you are
hitting it.  If there is an existing partition table then it will
present it to you and allow you to use it.  If the existing partition
table indicates RAID or LVM partitions then it will allow you to use
those too.  Which means that if you *don't* want to use them that you
need to do some more work to avoid the pre-existing ones and to set up
new ones.

There is an option to configure the LVM manager.  From memory I
believe you need to select that option and then delete and deactivate
LVM if you don't want to us it.  Same thing for RAID.  I remember this
being slightly confusing if you are wanting to install freshly upon it
because it shows you the previous pre-existing configuration.  But
many people want that because they have other operating systems on the
disk and they want to partition it using the other system first and
then install and this behavior allows them to keep the other systems
alive and intact and install around them.

If this is too much hassle then you can always zero out the partition
table first and then install after having zeroed the table.  (Or disk,
but that would take a while.)  Once the table is zeroed then you can
partition fresh.

However some things like RAID and LVM have signatures on disk and if
the partition turns out to be the same as before then the signature
will appear as valid and it may come alive again anyway.  Both lvm and
mdadm have ways to zero out their own respective signatures.

So, I plugged in a external USB hard drive and tried to install on
it. It refused to write on either drive until I did something with
the left over logical volume, mentioning it by name.
This is quite vague.  I can't determine what you did.  But if you will
pardon my bluntness I am certain that you selected a secure shred
erase option and that isn't what you were wanting to do.

As I write this email I am doing a partition erase on the logical
volume.
I think a data erase is where it actually writes random data over the
partition to securely void any previous data there.  That usually
isn't needed.

I tried selecting 'do not use this device' , which to me should have
been sufficient, but it wasn't.
No.  Because then the disk space would remain consumed.  You want to
free up the disk space and reallocate it to new logical volumes.  You
can't just leave it consumed but not used.

Erase is going slowly. At the rate the patience bar is growing, it
should be finished in another 12 hrs or so.
Sounds about right for the secure data shred option.  But you should
not need to do that.

Is there a better way? What should I try if it still claims my
architecture is not available on my mirror? etc.  Ideas?
I myself would simply look very carefully at the options at
installation time and adjust whatever is showing me.  For example if
it has lvm partitions already then I would "configure the lvm manager"
and remove the previous partitions and create new ones as I desire
them.  It is less hassle than trying to zero the disk.  You can also
unconfigure completely the lvm manager.  Back it all out.  Then reboot
and start the installation fresh.  With the lvm de-configured it won't
be pre-existing anymore.

Unfortunately my main debug VM system is Sid and my VM system is quite
broken at the moment due to Sid breakage.  So I can't just jump in and
create a VM and walk through the debian-installer in order to review
what needs to be done.  At the moment I would actually need to set up
real hardware to do that and then I couldn't take screenshots.  And it
would take much longer and no time to do it.  But I know that it is
possible.

Bob
Bob,

Thanks for the long response. I have a functioning Wheezy on the computer now, but issues remain. I think my experience and the way I got through it is not something I want to repeat, so I would like to revisit it issue when I have collected my wits. I actually made many more attempts to install than I mentioned. I think (preliminary theory) the original hard disk in the refurb was defective. What I reported was results I was getting was results after I replaced that HD with a new one. But I made the mistake of simply trying to do an install. I was puzzled because the running debootstrap, or whatever is the software running during a network install, seemed to already 'know' that this system has 'LVM' on it. How? I have removed the HD that was used. Rather than stopping and researching, I did want LVM, per your comments about it as a way of avoiding issues to device naming, so I pushed on. The options offered by the partitioner did NOT include an option to remove the LVM or an option to replace it with a new empty one. It complained that I had not created a root partition, whatever I tried. At one point I removed the external HD, which had been corrupted with false information about a supposed LVM and cleaned it of this corruption on a different computer. The problem with the architecture being unavailable was fixed by doing garbage collection on my approx server. My rapid fire attempts of Wheezy installs overloaded it. (garbage collection is normally done over night) But I really don't know what I did that was effective in getting rid of the old LVM, so when I did get to the point of not seeing LVM on either disk when doing the partitioning part of install, I decided to avoid LVM this time and the current condition of the computer is a single ordinary, old style partition, not LVM. I think the option to remove a pre-existing LVM is not there, or is hidden, or very cleverly disguised. So, the computer is there. It is working, but was purchased with a view to having an extra computer that I could use in case something bad happens with my main computer, or the one that I use for a print server (because it has a Centronics printer port for my old laser printer) I need to start soon preparing for the release of Jessie, but I probably won't make much progress in the near term.

Good luck on fixing your VM. It, and you, have solved many problems for puzzled users.

Thanks,
Paul



Reply to: