Re: [gentoo-user] Mounting Question...

2007-12-23 Thread Peter Humphrey
On Friday 21 December 2007 18:46:52 Mick wrote:

 The problem with some distros installation scripts is that they are
 trying to be too clever for their own good.  As a result they some times
 behave like MS Windows and unless you whip them into submission they
 could trash your system!

Exactly the cause of my nervousness.

 What I always do is to create the partitions and LV that I want and then
 instruct the distro in question to install itself in there.  Of course if
 the distro in question does not have an LVM compatible kernel then you'll
 need to install it using a different medium (with a kernel that has all
 the necessary drivers) and untar the distro's fs into your partitions of
 choice. 

So much simpler just to manage my own partitions. Good for the confidence, 
too, not to mention the feeling of knowing what's going on.

The rest of the world will no doubt continue to keep up with all the latest 
developments, but in this respect I prefer to remain a dinosaur. (They did, 
after all, rule the world for hundreds of millions of years, even if 
they're not still around today - thousands of times longer than we can yet 
claim.)

-- 
Rgds
Peter
-- 
[EMAIL PROTECTED] mailing list



Re: [gentoo-user] Mounting Question...

2007-12-22 Thread Dirk Heinrichs
Am Donnerstag, 20. Dezember 2007 schrieb Benjamen R. Meyer:

 I don't like using NFS much...guess I'll have to change that as I would
 like to centralize my server as a one-stop shop for usernames and
 passwords for the few systems on my network - server, desktop, and a
 laptop at present, but there will also be a few others shortly too. The
 laptop runs Windows 2k, so it'll just auth against Samba...any how...to
 get back to this issue...

Did you think about using OpenAFS?

 I haven't played with LVM yet. It's been something that's intrigued me,
 but I haven't ever researched it much to play with it. What you guys
 propose above and in this thread is quite interesting, so I'll follow up
 with this question:

 Right now I have the server configured per drives as follows:

 /dev/hda1   /    3.8 GB   4096.19 MB
 /dev/hda2   /home   15.0 GB  15356.60 MB
 /dev/hda3   SWAP 2.6 GB   2665.00 MB
 /dev/hda4   /usr/local   4.9 GB   5255.96 MB

 /dev/hdb1   EMPTY   66.3 GB  67875.02 MB
 /dev/hdb2   /var/tmp28.0 GB  30721.43 MB
 /dev/hdb3   /usr/portage47.0 GB  51202.37 MB
 /dev/hdb4   SWAP10.0 GB  10240.48 MB

Having the output of df would help a lot, because it shows how much space is 
already occupied on each filesystem. What about /usr/portage? If you have a 
broadband internet connection you don't need to care about it.

 It's only got a 192 MB of RAM - a PII/233, so I'm giving it generous
 swap space. (My desktop is an AMD64 with  a gig of RAM.) I seem to have
 a sizable partition free (hdb1), so this just might work - but how would
 you guys propose I transition from the above setup to an LVM setup? All
 partitions are currently ext3 (my preferred fs for linux).

Hmm, looks like hdb1 has enough space for all of hda. So you could just boot 
into a rescue CD (my recommendation: GRML), copy the stuff over, eventually  
revise fstab on hdb1 and boot from this partition (to make sure everything 
still wortks as before), then boot back into GRML and repartition hda and 
create logical volumes (as per my first reply), copy the stuff back, together 
with the remaining stuff from hdb, then repartition hdb and add it to the 
volume group.

If you want a more detailed description of the steps above, you can mail me 
directly.

Bye...

Dirk


signature.asc
Description: This is a digitally signed message part.


Re: [gentoo-user] Mounting Question...

2007-12-21 Thread Galevsky
On Dec 20, 2007 10:31 PM, Mick [EMAIL PROTECTED] wrote:
 Unlike commonly perceived wisdom I don't think that LVM is a panacea for all
 ills, or a necessity as such.  It is however bloody convenient, especially on
 a growing fs.  A server that is not expected to change much in size, probably
 does not need it.  On the other hand some servers (file, mail, news servers)
 are bound to continue to accumulate data and their fs will increase in time.
 I would argue that the former type of server can happily live in a few primary
 partitions + 1 extended with a number of logical partitions, if you are going
 for a multi-partitioned scheme, while the latter type of server will greatly
 benefit from LVM.  Of course, if hard drive redundancy is necessary, then I
 can't see how you could live without LVM + RAID.

I understand you on LVM is not a must for very stable servers, but
since I can't see any good reason not to use LVM,  I see no reason to
limit your abilities to extended partitions. We have the opportunity
to be more flexible with LVM, why should we not get it ? To loose the
ability to extend a partition by adding a new HD without any pain ? I
mean, if you don't know how to use it, I understand that you may skip
installing a LVM system, but when you did it once, I see no reason to
install your new systems without. So, I am interested in your advice
about LVM is not the universal solution for partitions management,
since I am sure I have something to learn from you experience.

Gal'
-- 
[EMAIL PROTECTED] mailing list



Re: [gentoo-user] Mounting Question...

2007-12-21 Thread Benjamen R. Meyer
Galevsky wrote:
 On Dec 20, 2007 10:31 PM, Mick [EMAIL PROTECTED] wrote:
 Unlike commonly perceived wisdom I don't think that LVM is a panacea for all
 ills, or a necessity as such.  It is however bloody convenient, especially on
 a growing fs.  A server that is not expected to change much in size, probably
 does not need it.  On the other hand some servers (file, mail, news servers)
 are bound to continue to accumulate data and their fs will increase in time.
 I would argue that the former type of server can happily live in a few 
 primary
 partitions + 1 extended with a number of logical partitions, if you are going
 for a multi-partitioned scheme, while the latter type of server will greatly
 benefit from LVM.  Of course, if hard drive redundancy is necessary, then I
 can't see how you could live without LVM + RAID.
 I understand you on LVM is not a must for very stable servers, but
 since I can't see any good reason not to use LVM,  I see no reason to
 limit your abilities to extended partitions. We have the opportunity
 to be more flexible with LVM, why should we not get it ? To loose the
 ability to extend a partition by adding a new HD without any pain ? I
 mean, if you don't know how to use it, I understand that you may skip
 installing a LVM system, but when you did it once, I see no reason to
 install your new systems without. So, I am interested in your advice
 about LVM is not the universal solution for partitions management,
 since I am sure I have something to learn from you experience.

Agreed. As I said in another e-mail on the list, I use to use extended
partitions - at one point I had about 10 or so partitions on a single
drive (3 primary, the rest from an extended partition). This worked well
under Windows 9x, but was a pain after moving to Linux. It wasn't that I
had mis-scoped the size of the data for those partitions, just that my
needs changed (mainly user related needs, not system related needs), and
managing extended partitions is a lot of work. I very much understand
LVM and what would do for me, and would very much like to hear why
simple extended partitions would be better for any scenario but the most
limited of scenarios where LVM was just not possible (e.g. the system
could not run a kernel that supported LVM; or RAM on the system was too
limited to support running LVM; etc.)...I'm not sure I agree that they
would be.

Ben
-- 
[EMAIL PROTECTED] mailing list



Re: [gentoo-user] Mounting Question...

2007-12-21 Thread Peter Humphrey
On Friday 21 December 2007 10:00:40 Galevsky wrote:

 I am interested in your advice [that] LVM is not the universal solution
 for partition management,  

In the case under discussion, namely a stable server, I wouldn't challenge 
any advice to use LVM, but I was using it until recently on this 
semi-experimental desktop box and found I was getting too nervous for 
comfort. From time to time I would be tempted to give another distribution 
a spin, and every time I did it was unable to recognise my existing 
partitions (and therefore leave them alone). The same was true of a couple 
of rescue CDs I tried - which of course meant I couldn't use them. That 
isn't a problem now, not since I installed a small rescue system on a spare 
disk in the same box.

So, for an unchanging system setup, by all means use LVM; for toy boxes it 
seems to me not to offer much advantage.

Incidentally, I have 4 GB RAM in this dual-246 box, so I've put /tmp into a 
tmpfs, which greatly speeds emerges. This is from /etc/fstab:
tmpfs   /tmptmpfs   nodev,nosuid,size=6g0 0

and this is from /etc/make.conf:
BUILD_PREFIX=/tmp/portage/build
PKG_TMPDIR=/tmp
PORTAGE_TMPDIR=/tmp
PORTAGE_TMPFS=/dev/shm

The disks get a holiday (except when compiling Open Office)   :-)

-- 
Rgds
Peter
-- 
[EMAIL PROTECTED] mailing list



Re: [gentoo-user] Mounting Question...

2007-12-21 Thread Mick
On Friday 21 December 2007, Benjamen R. Meyer wrote:
 Galevsky wrote:
  On Dec 20, 2007 10:31 PM, Mick [EMAIL PROTECTED] wrote:
  Unlike commonly perceived wisdom I don't think that LVM is a panacea for
  all ills, or a necessity as such.  It is however bloody convenient,
  especially on a growing fs.  A server that is not expected to change
  much in size, probably does not need it.  On the other hand some servers
  (file, mail, news servers) are bound to continue to accumulate data and
  their fs will increase in time. I would argue that the former type of
  server can happily live in a few primary partitions + 1 extended with a
  number of logical partitions, if you are going for a multi-partitioned
  scheme, while the latter type of server will greatly benefit from LVM. 
  Of course, if hard drive redundancy is necessary, then I can't see how
  you could live without LVM + RAID.
 
  I understand you on LVM is not a must for very stable servers, but
  since I can't see any good reason not to use LVM,  I see no reason to
  limit your abilities to extended partitions. We have the opportunity
  to be more flexible with LVM, why should we not get it ? To loose the
  ability to extend a partition by adding a new HD without any pain ? I
  mean, if you don't know how to use it, I understand that you may skip
  installing a LVM system, but when you did it once, I see no reason to
  install your new systems without. So, I am interested in your advice
  about LVM is not the universal solution for partitions management,
  since I am sure I have something to learn from you experience.

 Agreed. As I said in another e-mail on the list, I use to use extended
 partitions - at one point I had about 10 or so partitions on a single
 drive (3 primary, the rest from an extended partition). This worked well
 under Windows 9x, but was a pain after moving to Linux. It wasn't that I
 had mis-scoped the size of the data for those partitions, just that my
 needs changed (mainly user related needs, not system related needs), and
 managing extended partitions is a lot of work. I very much understand
 LVM and what would do for me, and would very much like to hear why
 simple extended partitions would be better for any scenario but the most
 limited of scenarios where LVM was just not possible (e.g. the system
 could not run a kernel that supported LVM; or RAM on the system was too
 limited to support running LVM; etc.)...I'm not sure I agree that they
 would be.

Guys, mine is not any precious experience that you could learn much from (I am 
sure others on this list have more valuable experience on this matter), but 
what I am saying is this:

If you have a stable, dedicated server which is NOT going to increase in fs 
size requirements, then a conventional non-LVM installation will do exactly 
what you need done, in a simpler fashion.  To define simpler in a server 
use case, I would say that anything that you do not absolutely need should 
not be installed (for basic security and maintainability reasons), including 
LVM kernel modules and what not.  On the other hand, installing and 
maintaining an LVM based fs is clearly not difficult and if you are uncertain 
about your current/future fs size requirements, then you're better off 
installing LVM and making use of the flexibility it offers.

BTW, if you're thinking of the flexibility of adding drives/partitions and 
extending LVG's at will, you should also consider that unless you're running 
a mirror RAID when any-one of your drives goes bang! you will lose all your 
VG data irrespective on which drive (PV) they reside.  Of course, you know 
this and you keep recent back ups of your data at all times, right?  ;-)

I can recall at least 4 server installations where I did not run LVM and I 
never had to increase the fs size (one of them has been running for more than 
3 years now and it fs is spread over two drives).  On the other hand a server 
I built less than two months ago has LVM and all data (but not its / ) is 
stored in LVs.  I already had to replace a drive on that machine which was 
suspect for an imminent failure.  A case of horses for courses.

Just my 2c's.
-- 
Regards,
Mick


signature.asc
Description: This is a digitally signed message part.


Re: [gentoo-user] Mounting Question...

2007-12-21 Thread Mick
On Friday 21 December 2007, Peter Humphrey wrote:
 On Friday 21 December 2007 10:00:40 Galevsky wrote:
  I am interested in your advice [that] LVM is not the universal solution
  for partition management,

 In the case under discussion, namely a stable server, I wouldn't challenge
 any advice to use LVM, but I was using it until recently on this
 semi-experimental desktop box and found I was getting too nervous for
 comfort. From time to time I would be tempted to give another distribution
 a spin, and every time I did it was unable to recognise my existing
 partitions (and therefore leave them alone). The same was true of a couple
 of rescue CDs I tried - which of course meant I couldn't use them. 

The problem with some distros installation scripts is that they are trying to 
be too clever for their own good.  As a result they some times behave like MS 
Windows and unless you whip them into submission they could trash your 
system!  What I always do is to create the partitions and LV that I want and 
then instruct the distro in question to install itself in there.  Of course 
if the distro in question does not have an LVM compatible kernel then you'll 
need to install it using a different medium (with a kernel that has all the 
necessary drivers) and untar the distro's fs into your partitions of choice.

 That 
 isn't a problem now, not since I installed a small rescue system on a spare
 disk in the same box.

 So, for an unchanging system setup, by all means use LVM; for toy boxes it
 seems to me not to offer much advantage.

 Incidentally, I have 4 GB RAM in this dual-246 box, so I've put /tmp into a
 tmpfs, which greatly speeds emerges. This is from /etc/fstab:
 tmpfs /tmptmpfs   nodev,nosuid,size=6g0 0

 and this is from /etc/make.conf:
 BUILD_PREFIX=/tmp/portage/build
 PKG_TMPDIR=/tmp
 PORTAGE_TMPDIR=/tmp
 PORTAGE_TMPFS=/dev/shm

 The disks get a holiday (except when compiling Open Office)   :-)

 --
 Rgds
 Peter



-- 
Regards,
Mick


signature.asc
Description: This is a digitally signed message part.


Re: [gentoo-user] Mounting Question...

2007-12-21 Thread Mark Kirkwood

Benjamen R. Meyer wrote:

I set up a server system a little while ago, and in performing updates
to portage it ran out of disk space as I didn't quite allow enough space
on the root partition (3.8 GB). As a result, I took a partition that I
had cleaned up (this was from a rebuild of a system that was a different
distro in the past) and moved over /usr/portage to it. It's a 47 GB
partition (as reported by df -h) and the system works fine.

I do realize that if the mount command got screwed up, I'd probably have
issues recovering the system, but that is that system.

I am now thinking of converting my desktop over to Gentoo as well, and
was wondering whether what I did above on the server was wise or not. I
will be using the server as the portage provider for my desktop too.
Otherwise, what is the recommended space to have available for the
portage tree in /usr/portage so I can have root as an appropriately
sized partition?

  

I'd recommend having a read of:

http://www.freebsd-howto.com/HOWTO/Filesystem-Layout-HOWTO

Now, although its a Freebsd resource, the ideas apply equally well to 
Linux (or UNIX for that matter - though you can skip where it discusses 
Freebsd partition and slice naming). In particular it discusses why 
separating /, /usr, /var, /tmp, /home is well worth doing - even tho it 
wasts a bit of space!


I used build systems with / including /usr and /var  but these days 
I do not make these part of / (for reasons covered in the article).


The downside is you end up with a lot of partitions and filesystems to 
figure out how to size - but you can use LVM make it a bit more 
forgiving if you need to resize them.


Cheers

Mark
--
[EMAIL PROTECTED] mailing list



[gentoo-user] Mounting Question...

2007-12-20 Thread Benjamen R. Meyer
I set up a server system a little while ago, and in performing updates
to portage it ran out of disk space as I didn't quite allow enough space
on the root partition (3.8 GB). As a result, I took a partition that I
had cleaned up (this was from a rebuild of a system that was a different
distro in the past) and moved over /usr/portage to it. It's a 47 GB
partition (as reported by df -h) and the system works fine.

I do realize that if the mount command got screwed up, I'd probably have
issues recovering the system, but that is that system.

I am now thinking of converting my desktop over to Gentoo as well, and
was wondering whether what I did above on the server was wise or not. I
will be using the server as the portage provider for my desktop too.
Otherwise, what is the recommended space to have available for the
portage tree in /usr/portage so I can have root as an appropriately
sized partition?

TIA,

Ben
-- 
[EMAIL PROTECTED] mailing list



Re: [gentoo-user] Mounting Question...

2007-12-20 Thread Galevsky
I won't answer you with a size since its mainly depends on your own
needs, but don't you know that solutions like lvm or evms provide lots
of flexibility to manage your HD resources ? I advise you to look at
lvm howto. It allows you to add/remove/move/enlarge your partitions as
you need in a truly painless way. To avoid major mounting problems
(can be done but with caution), let /boot and / outside lvm, and put
the others in logical volumes.

Gal'


On Dec 20, 2007 10:50 AM, Benjamen R. Meyer [EMAIL PROTECTED] wrote:
 I set up a server system a little while ago, and in performing updates
 to portage it ran out of disk space as I didn't quite allow enough space
 on the root partition (3.8 GB). As a result, I took a partition that I
 had cleaned up (this was from a rebuild of a system that was a different
 distro in the past) and moved over /usr/portage to it. It's a 47 GB
 partition (as reported by df -h) and the system works fine.

 I do realize that if the mount command got screwed up, I'd probably have
 issues recovering the system, but that is that system.

 I am now thinking of converting my desktop over to Gentoo as well, and
 was wondering whether what I did above on the server was wise or not. I
 will be using the server as the portage provider for my desktop too.
 Otherwise, what is the recommended space to have available for the
 portage tree in /usr/portage so I can have root as an appropriately
 sized partition?

 TIA,

 Ben
 --
 [EMAIL PROTECTED] mailing list


-- 
[EMAIL PROTECTED] mailing list



Re: [gentoo-user] Mounting Question...

2007-12-20 Thread Dirk Heinrichs
On Thursday 20 December 2007 10:50:33 Benjamen R. Meyer wrote:

 I set up a server system a little while ago, and in performing updates
 to portage it ran out of disk space as I didn't quite allow enough space
 on the root partition (3.8 GB).

That's way too much. 256M is enough.

 As a result, I took a partition that I had cleaned up (this was from a
 rebuild of a system that was a different
 distro in the past) and moved over /usr/portage to it. It's a 47 GB
 partition (as reported by df -h) and the system works fine.

 I do realize that if the mount command got screwed up, I'd probably have
 issues recovering the system, but that is that system.

 I am now thinking of converting my desktop over to Gentoo as well, and
 was wondering whether what I did above on the server was wise or not.

I think it is not. You'll undoubtedly get different answers about this, but 
IMHO it is best (regardless what kind of system) to use small, special purpose 
logical volumes. This way you can add space when needed, use the filesystem 
that fits best for the kind of data you store on this volume and have a certain 
degree of safety against volume corruption.

Here is what I would recommend for a normal linux system:

[hs]da1: /boot, 64M, ext2
[hs]da2: /, 256M, ext3 or xfs
[hs]da3: LVM

Then, create a volume group spawning [hs]da3 with name vg00 (you can choose the 
name freely) and create logical volumes inside:

/dev/vg00/swap: size as needed, swapfs # can be omitted if enough RAM
/dev/vg00/usr: /usr, 2-5G (dep. on number of pkgs), ext3 or xfs
/dev/vg00/var: /var, 512M-1G, ext3 or xfs

For /home, I prefer to have one LV per user, like /dev/vg00/john_doe, 
/dev/vg00/jane_doe and have the kernel automounter mount them on demand (at 
login time).

 I will be using the server as the portage provider for my desktop too.
 Otherwise, what is the recommended space to have available for the
 portage tree in /usr/portage so I can have root as an appropriately
 sized partition?

Here again, I use the kernel automounter to mount three different LVs under 
/gentoo when needed: /dev/vg00/build (5.5G to be able to build OO.org), 
/dev/vg00/distfiles for the source packages and /dev/vg00/overlays for 
overlays, incl. the portage tree.

On the desktop machine, you should be able to mount distfiles and overlays from 
the server via NFS. The build volume I would leave locally on the desktop to 
get faster build times (unless your network connection to your server is faster 
than harddisc access).

HTH...

Dirk

Re: [gentoo-user] Mounting Question...

2007-12-20 Thread Neil Bothwick
On Thu, 20 Dec 2007 15:12:17 +0100, Dirk Heinrichs wrote:

 Then, create a volume group spawning [hs]da3 with name vg00 (you can
 choose the name freely) and create logical volumes inside:

I'd use a less generic name, otherwise you'll have problems if the
computer fails and you try to connect the disk to another computer that
has a vg00 volume group. I generally use a name related to the computer's
hostname, which avoids conflicts.


-- 
Neil Bothwick

.sig a .sog of sixpence.


signature.asc
Description: PGP signature


Re: [gentoo-user] Mounting Question...

2007-12-20 Thread Dan Farrell
On Thu, 20 Dec 2007 09:50:33 +
Benjamen R. Meyer [EMAIL PROTECTED] wrote:

 I set up a server system a little while ago, and in performing updates
 to portage it ran out of disk space as I didn't quite allow enough
 space on the root partition (3.8 GB). As a result, I took a partition
 that I had cleaned up (this was from a rebuild of a system that was a
 different distro in the past) and moved over /usr/portage to it. It's
 a 47 GB partition (as reported by df -h) and the system works fine.

47 Gigs is quite a lot of space for the portage tree.  The ebuilds
take up a few hundred megabytes, and the distfiles generally fill up a
few gigs by the time you get everything installed (i recommend you keep
them around, if you plan to share portage with others).  The amount of
space you provide is overkill -- but more importantly, I worry that you
may need to reclaim some of that 47GB to use for your root partition,
as 3.8GB isn't a particularly large amount for this, especially if you
have /var on that partition as well.  

I generally go 5 or 10 gigs for a root partition, the latter being more
appropriate for a general purpose graphical workstation.  My shared
portage tree has been deployed for quite a while now and is about 5.4
gigs, with roughly 5G of that accounted for by distfiles.  

 I am now thinking of converting my desktop over to Gentoo as well, and
 was wondering whether what I did above on the server was wise or not.
 I will be using the server as the portage provider for my desktop too.
 Otherwise, what is the recommended space to have available for the
 portage tree in /usr/portage so I can have root as an appropriately
 sized partition?

I believe that sharing portage between computers is wise whenever the
client is guaranteed to have access to the tree when new programs are
required.  In other words, it works great for desktops.  The
configuration might come back to haunt you if you are off your own
network and need to install a new program on your laptop.  
-- 
[EMAIL PROTECTED] mailing list



Re: [gentoo-user] Mounting Question...

2007-12-20 Thread Benjamen R. Meyer
Dirk Heinrichs wrote:
 On Thursday 20 December 2007 10:50:33 Benjamen R. Meyer wrote:
 I set up a server system a little while ago, and in performing updates
 to portage it ran out of disk space as I didn't quite allow enough space
 on the root partition (3.8 GB).
 That's way too much. 256M is enough.

/ is the primary drive for the OS; I typically only off-load to other
partitions for user stuff. On the server, I initially only offloaded
/home and /usr/local; but in the crisis of the out of diskspace issue,
I ended up also offloading /var/tmp and /usr/portage.

 As a result, I took a partition that I had cleaned up (this was from a
 rebuild of a system that was a different
 distro in the past) and moved over /usr/portage to it. It's a 47 GB
 partition (as reported by df -h) and the system works fine.
 I do realize that if the mount command got screwed up, I'd probably have
 issues recovering the system, but that is that system.
 I am now thinking of converting my desktop over to Gentoo as well, and
 was wondering whether what I did above on the server was wise or not.
 I think it is not. You'll undoubtedly get different answers about this,
 but IMHO it is best (regardless what kind of system) to use small,
 special purpose logical volumes. This way you can add space when needed,
 use the filesystem that fits best for the kind of data you store on this
 volume and have a certain degree of safety against volume corruption.
 Here is what I would recommend for a normal linux system:
 [hs]da1: /boot, 64M, ext2
 [hs]da2: /, 256M, ext3 or xfs
 [hs]da3: LVM
 Then, create a volume group spawning [hs]da3 with name vg00 (you can
 choose the name freely) and create logical volumes inside:
 /dev/vg00/swap: size as needed, swapfs # can be omitted if enough RAM
 /dev/vg00/usr: /usr, 2-5G (dep. on number of pkgs), ext3 or xfs
 /dev/vg00/var: /var, 512M-1G, ext3 or xfs
 For /home, I prefer to have one LV per user, like /dev/vg00/john_doe,
 /dev/vg00/jane_doe and have the kernel automounter mount them on demand
 (at login time).
 I will be using the server as the portage provider for my desktop too.
 Otherwise, what is the recommended space to have available for the
 portage tree in /usr/portage so I can have root as an appropriately
 sized partition?
 Here again, I use the kernel automounter to mount three different LVs
 under /gentoo when needed: /dev/vg00/build (5.5G to be able to build
 OO.org), /dev/vg00/distfiles for the source packages and
 /dev/vg00/overlays for overlays, incl. the portage tree.
 On the desktop machine, you should be able to mount distfiles and
 overlays from the server via NFS. The build volume I would leave locally
 on the desktop to get faster build times (unless your network connection
 to your server is faster than harddisc access).

I don't like using NFS much...guess I'll have to change that as I would
like to centralize my server as a one-stop shop for usernames and
passwords for the few systems on my network - server, desktop, and a
laptop at present, but there will also be a few others shortly too. The
laptop runs Windows 2k, so it'll just auth against Samba...any how...to
get back to this issue...

I haven't played with LVM yet. It's been something that's intrigued me,
but I haven't ever researched it much to play with it. What you guys
propose above and in this thread is quite interesting, so I'll follow up
with this question:

Right now I have the server configured per drives as follows:

/dev/hda1   /3.8 GB   4096.19 MB
/dev/hda2   /home   15.0 GB  15356.60 MB
/dev/hda3   SWAP 2.6 GB   2665.00 MB
/dev/hda4   /usr/local   4.9 GB   5255.96 MB

/dev/hdb1   EMPTY   66.3 GB  67875.02 MB
/dev/hdb2   /var/tmp28.0 GB  30721.43 MB
/dev/hdb3   /usr/portage47.0 GB  51202.37 MB
/dev/hdb4   SWAP10.0 GB  10240.48 MB

It's only got a 192 MB of RAM - a PII/233, so I'm giving it generous
swap space. (My desktop is an AMD64 with  a gig of RAM.) I seem to have
a sizable partition free (hdb1), so this just might work - but how would
you guys propose I transition from the above setup to an LVM setup? All
partitions are currently ext3 (my preferred fs for linux).

I don't think I'd be able to do that on my desktop right now...namely in
that rebuilding it from Slackware to Gentoo is going to be trying
enough, but I think I can manage it - namely from the side of downtime,
but I'd also like to try to fully utilize the AMD64 in the system -
meaning 64-bit where possible. Any how...for now, I'd like to hear about
the LVM conversion for the server; I'll bring up the other issues later
in different threads when I have the time to address them, but the LVM
stuff is intriguing enough that I might be able to squeeze it in in
short order if I can do it without risking data, or having to rebuild
the system.

Thanks,

Ben
-- 
[EMAIL PROTECTED] mailing 

Re: [gentoo-user] Mounting Question...

2007-12-20 Thread Mick
On Thursday 20 December 2007, Neil Bothwick wrote:
 On Thu, 20 Dec 2007 15:12:17 +0100, Dirk Heinrichs wrote:
  Then, create a volume group spawning [hs]da3 with name vg00 (you can
  choose the name freely) and create logical volumes inside:

 I'd use a less generic name, otherwise you'll have problems if the
 computer fails and you try to connect the disk to another computer that
 has a vg00 volume group. I generally use a name related to the computer's
 hostname, which avoids conflicts.

I can already see that this thread is going to run, and run, and ...  :)

These days most people do not have a separate /boot partition as has already 
been mentioned.  Depending on the size of your disk and your need for a swap 
partition you may want to have it at the beginning of a partition, or for 
larger disks in the middle.  At the beginning you get faster read/write and 
in the middle you get faster access (I'm splitting hairs here, but it's fun 
anyway).  Certain partitions (if you decide to go for multi-partition scheme) 
like /var/tmp, /tmp, /usr will benefit being at the beginning of the disk. 
Others (e.g. /root, /mnt, /sbin less so).

Unlike commonly perceived wisdom I don't think that LVM is a panacea for all 
ills, or a necessity as such.  It is however bloody convenient, especially on 
a growing fs.  A server that is not expected to change much in size, probably 
does not need it.  On the other hand some servers (file, mail, news servers) 
are bound to continue to accumulate data and their fs will increase in time.  
I would argue that the former type of server can happily live in a few primary 
partitions + 1 extended with a number of logical partitions, if you are going 
for a multi-partitioned scheme, while the latter type of server will greatly 
benefit from LVM.  Of course, if hard drive redundancy is necessary, then I 
can't see how you could live without LVM + RAID.

With regards to your 47G /usr/portage partition I think that it is a waste of 
space.  It won't harm you other than the fact that the 3.8G OS partition is 
in all likelihood too small.  This is what I would do: tar the contents 
of /usr/portage elsewhere (even in the 3.8G partition - it should fit if you 
clear any cruft and, or use bzip).  Delete the 47G partition and use gparted 
to enlarge the 3.8G partition to say, 8-10G.  Then create a new partition say 
another 8-10G for /usr/portage.  Then create anymore separate partitions you 
may need (for /home and what have you). mkfs as required, modify 
your /etc/fstab and move your data in your respective new partitions.  If you 
think your fs is/are going to grow use LVM instead, otherwise primaries and 
if you need more than 4 then (extended + logical).

Just my 2c's.
-- 
Regards,
Mick


signature.asc
Description: This is a digitally signed message part.


Re: [gentoo-user] Mounting Question...

2007-12-20 Thread Dale
Mick wrote:
 SNIP

 With regards to your 47G /usr/portage partition I think that it is a waste of 
 space.  It won't harm you other than the fact that the 3.8G OS partition is 
 in all likelihood too small.  This is what I would do: tar the contents 
 of /usr/portage elsewhere (even in the 3.8G partition - it should fit if you 
 clear any cruft and, or use bzip).  Delete the 47G partition and use gparted 
 to enlarge the 3.8G partition to say, 8-10G.  Then create a new partition say 
 another 8-10G for /usr/portage.  Then create anymore separate partitions you 
 may need (for /home and what have you). mkfs as required, modify 
 your /etc/fstab and move your data in your respective new partitions.  If you 
 think your fs is/are going to grow use LVM instead, otherwise primaries and 
 if you need more than 4 then (extended + logical).

 Just my 2c's.
   

Well, I'm no expert but this has worked for me and this is a 4 or 5 year
old install.  Your mileage may vary.  From cfdisk:


hda1  Boot  Primary Linux
ext2  200.25
hda5Logical
Linux   1999.88
hda6Logical Linux
ReiserFS1.76
hda7Logical Linux
ReiserFS 4999.94
hda8Logical Linux
ReiserFS 4999.94
hda9Logical Linux
ReiserFS49764.56

This is from df:

[EMAIL PROTECTED] / # df
Filesystem   1K-blocks  Used Available Use%
Mounted on
/dev/hda6 19530340   4842172  1468816825%
  /
/dev/hda1   189339 19382  160181   
11% /boot
/dev/hda7  4882532   2832424   2050108  59%
/usr/portage
/dev/hda8  4882532   1597144   3285388  33%
/home
/dev/hdb1 78145768  13720248 6442552018%
/data

[EMAIL PROTECTED] / #

As you can see, I have plenty of space available for future additions,
like a space hogging KDE 4.0.  :-)  The fullest one is /usr/portage
which I clean up on occasion with eclean.  If I ever change them around
again, I will put /var on a separate partition but other than that, it
works pretty well.  May make root smaller then as well.

A lot of this depends on what you are doing with the box tho.  It's just
something you have to sort of work out as you go which may be why some
recommend EVMS or LVM.  I have read up on it but just never got up the
nerve to try it yet.  This is a desktop mostly used to surf the net and
run foldingathome on.

Hope this helps tho.

Dale

:-)  :-)  :-) 
-- 
[EMAIL PROTECTED] mailing list



Re: [gentoo-user] Mounting Question...

2007-12-20 Thread Benjamen R. Meyer
Dale wrote:
 Mick wrote:
 SNIP
 With regards to your 47G /usr/portage partition I think that it is a waste 
 of 
 space.  It won't harm you other than the fact that the 3.8G OS partition is 
 in all likelihood too small.  This is what I would do: tar the contents 
 of /usr/portage elsewhere (even in the 3.8G partition - it should fit if you 
 clear any cruft and, or use bzip).  Delete the 47G partition and use gparted 
 to enlarge the 3.8G partition to say, 8-10G.  Then create a new partition 
 say 
 another 8-10G for /usr/portage.  Then create anymore separate partitions you 
 may need (for /home and what have you). mkfs as required, modify 
 your /etc/fstab and move your data in your respective new partitions.  If 
 you 
 think your fs is/are going to grow use LVM instead, otherwise primaries and 
 if you need more than 4 then (extended + logical).
 Well, I'm no expert but this has worked for me and this is a 4 or 5 year
 old install.  Your mileage may vary.  From cfdisk:
snip
 As you can see, I have plenty of space available for future additions,
 like a space hogging KDE 4.0.  :-)  The fullest one is /usr/portage
 which I clean up on occasion with eclean.  If I ever change them around
 again, I will put /var on a separate partition but other than that, it
 works pretty well.  May make root smaller then as well.
 A lot of this depends on what you are doing with the box tho.  It's just
 something you have to sort of work out as you go which may be why some
 recommend EVMS or LVM.  I have read up on it but just never got up the
 nerve to try it yet.  This is a desktop mostly used to surf the net and
 run foldingathome on.
 Hope this helps tho.

Thanks for the info guys. Yeah - the server has been pretty steady. I
use to run it on a P90 with an 8.4 GB (7.6 formatted) hard drive running
Slackware and just upgraded to the P2 with Gentoo, namely so I can keep
it up to date more. I run Gentoo at work, but the firewall prevents me
from getting portage updates there as they block RSYNC and FTP, and the
HTTP is authenticated which causes me a lot of pain under *nix. So in
some respects I am pretty new to some of this stuff per Gentoo.

LVM is certainly not out of the question, I just don't have the time to
rebuild the system again - especially since I just built it. So I'd need
a path to getting to it.

As per the the suggestion of blasting away the 47 GB partition - I'm not
sure that's an option. I got away from using Logical partitions a long
time ago after I moved to Linux as I found them to be too problematic -
I'd never have enough space on the partition I needed space on and to
rework it to have enough would require moving around others too. And, as
you can see from my other e-mail, I already have 4 primary partitions on
each drive (swap included); so I would certainly go to LVM instead of
logical partitions.

That said, the system itself won't change much, but the current drive
layout is probably not the best for where space needs to really be. So I
really am open to changing it, but need to do so on the fly with a few
reboots and (most importantly) without reinstalling. I do realize Linux
makes it pretty easy to move around from partition to partition, which I
have done, just not sure how LVM plays into it - thus my other e-mail
asking about a path to getting there. (FYI - I did check and LVM2's
device-mapper is enabled in the kernel, so it should be pretty straight
forward.)

Thanks,

Ben
-- 
[EMAIL PROTECTED] mailing list