Re: Any caveats moving root filesystem to LVM?
On Jan 21, 2005, at 3:40 PM, David Boyes wrote: What I use is: 1) / as ext2 or ext3 (depending on distribution and platform -- usually ext2 on zSeries) 2) all other filesystems as ext3 3) if a file system needs to be bigger than a physical volume, then use LVM and create ext3 filesystems on the logical volumes created by LVM. That works on pretty-much all flavors of Linux and all platforms, and if something goes horribly wrong, then diagnostics are pretty straightforward, and I can usually get the systems up to the point that you can at least try to fix things w/o a rescue system. In general, I agree with David, with the following addenda: Sometimes it's nice to separate /boot out into its own partition, holding basically the kernel, the initrd (if applicable), and the IPL code at the start of the partition. This can easily be ext2 because it's only 20-30 MB, usually (depending on how many old kernel versions you want to keep) and therefore is very quick to check. Also, some people have reported success with ReiserFS on LVM partitions. ReiserFS is great if you have lots of little files. However, we encountered a situation where we got nasty data corruption with it under extremely heavy load. I do not know if this has been fixed in more recent versions of it, but it was enough to scare me off ReiserFS on S/390. Ext3 is not great performance but it is very reliable and has what I think is a great advantage that an ext3 fs is an ext2 fs plus stuff, so you can work with it, if you need to, as if it were an ext2 filesystem, which makes recovery somewhat easier (not nearly as easy, of course, as having decent backups in the first place!). Adam -- For LINUX-390 subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390
Re: Any caveats moving root filesystem to LVM?
You can always increase the size of the LV dynamically. That applies to all filesystems. The problem is that ext2/3 will not allow you to add space (using resize2fs) unless the FS is unmounted. With Reiser though, you can use the resize_reiserfs command on a mounted filesystem. We used it successfully many times. The sequence is: pvcreate (physical volume) vgextend (to add the volume to the VG) lvextend (to increase the size of the volume) resize_reiserfs (to add the free space to the FS) If you give resize_reiserfs no parameters other than the filesystem name, it's smart enough to fill the remaining space on the volume. REDUCING the size of a filesystem is also possible, but I've only done it once on Linux/Intel, and it was much more complicated than increasing it. It also took a LONG time. Far longer than adding space. When you add striping to the mix, you lose the ability to do lvextend, since the number of stripes has to match the number of physical volumes in the LV. > -Original Message- > From: Linux on 390 Port [mailto:[EMAIL PROTECTED] Behalf Of > Romanowski, John (OFT) > Sent: Monday, January 24, 2005 10:40 AM > To: LINUX-390@VM.MARIST.EDU > Subject: Re: [LINUX-390] Any caveats moving root filesystem to LVM? > > > How come in LVM without striping, with Reiser FS you can resize > filesystems without unmounting them? I'm just curious. > > -Original Message- > From: Linux on 390 Port [mailto:[EMAIL PROTECTED] On Behalf Of > Hall, Ken (IDS DCS PE) > Sent: Monday, January 24, 2005 10:33 AM > To: LINUX-390@VM.MARIST.EDU > Subject: Re: Any caveats moving root filesystem to LVM? > > If you use LVM without striping, and Reiser FS on top of > that, you don't > even have to unmount the filesystems to resize them. > > Striping takes away ALL of this advantage. We ran some tests with > striped vs. non-striped filesystems, and generally got better results > with striping, but I don't have numbers anymore. > > Regardless, I'm not sure if you would get much benefit out of striping > the root FS, since the benefit of striping is in > parallelizing I/O. The > files in the root FS tend to be small, and are either read > infrequently > (at boot time), or read SO frequently that they tend to stay in the > buffer cache. Write activity against the root FS should be EXTREMELY > limited by design (although we had some products that > stubbornly insist > on putting things like dumps in /). > > -- > For LINUX-390 subscribe / signoff / archive access instructions, > send email to [EMAIL PROTECTED] with the message: INFO > LINUX-390 or visit > http://www.marist.edu/htbin/wlvindex?LINUX-390 > If you are not an intended recipient of this e-mail, please notify the sender, delete it and do not read, act upon, print, disclose, copy, retain or redistribute it. Click here for important additional terms relating to this e-mail. http://www.ml.com/email_terms/ -- For LINUX-390 subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390
Re: Any caveats moving root filesystem to LVM?
How come in LVM without striping, with Reiser FS you can resize filesystems without unmounting them? I'm just curious. -Original Message- From: Linux on 390 Port [mailto:[EMAIL PROTECTED] On Behalf Of Hall, Ken (IDS DCS PE) Sent: Monday, January 24, 2005 10:33 AM To: LINUX-390@VM.MARIST.EDU Subject: Re: Any caveats moving root filesystem to LVM? If you use LVM without striping, and Reiser FS on top of that, you don't even have to unmount the filesystems to resize them. Striping takes away ALL of this advantage. We ran some tests with striped vs. non-striped filesystems, and generally got better results with striping, but I don't have numbers anymore. Regardless, I'm not sure if you would get much benefit out of striping the root FS, since the benefit of striping is in parallelizing I/O. The files in the root FS tend to be small, and are either read infrequently (at boot time), or read SO frequently that they tend to stay in the buffer cache. Write activity against the root FS should be EXTREMELY limited by design (although we had some products that stubbornly insist on putting things like dumps in /). -- For LINUX-390 subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390
Re: Any caveats moving root filesystem to LVM?
If you use LVM without striping, and Reiser FS on top of that, you don't even have to unmount the filesystems to resize them. Striping takes away ALL of this advantage. We ran some tests with striped vs. non-striped filesystems, and generally got better results with striping, but I don't have numbers anymore. Regardless, I'm not sure if you would get much benefit out of striping the root FS, since the benefit of striping is in parallelizing I/O. The files in the root FS tend to be small, and are either read infrequently (at boot time), or read SO frequently that they tend to stay in the buffer cache. Write activity against the root FS should be EXTREMELY limited by design (although we had some products that stubbornly insist on putting things like dumps in /). > -Original Message- > From: Linux on 390 Port [mailto:[EMAIL PROTECTED] Behalf Of > Mrohs, Ray > Sent: Monday, January 24, 2005 9:41 AM > To: LINUX-390@VM.MARIST.EDU > Subject: Re: [LINUX-390] Any caveats moving root filesystem to LVM? > > > Striping does give you more paths into storage, but I haven't seen any > performance studies of striped vs. non-striped LVM disks. If > you have fast > hardware and FICON, the advantage might not be that great. > With striping, you > lose the benefit of adding or removing physical volumes > dynamically in the volume > group (well, you still have to umount the file system > briefly), which means you > have to plan your file system growth really well, or take a > chunk of down time to > dump, resize, and reload whenever you need more space. We use > the non-striped > variety. > > Ray Mrohs > Energy Information Administration > U.S. Department of Energy > > > -Original Message- > From: David Andrews [mailto:[EMAIL PROTECTED] > Sent: Monday, January 24, 2005 8:50 AM > To: LINUX-390@VM.MARIST.EDU > Subject: Re: Any caveats moving root filesystem to LVM? > > > On Fri, 2005-01-21 at 16:02 -0500, Mrohs, Ray wrote: > > My rule-of-thumb is to only use LVM when it's necessary, as > in providing more > > file system space than one minidisk can provide. > > There is also some striping value, no? (At least until Linux supports > PAV.) > > -- > David Andrews > A. Duda and Sons, Inc. > [EMAIL PROTECTED] > > -- > For LINUX-390 subscribe / signoff / archive access instructions, > send email to [EMAIL PROTECTED] with the message: INFO > LINUX-390 or visit > http://www.marist.edu/htbin/wlvindex?LINUX-390 > > -- > For LINUX-390 subscribe / signoff / archive access instructions, > send email to [EMAIL PROTECTED] with the message: INFO > LINUX-390 or visit > http://www.marist.edu/htbin/wlvindex?LINUX-390 > If you are not an intended recipient of this e-mail, please notify the sender, delete it and do not read, act upon, print, disclose, copy, retain or redistribute it. Click here for important additional terms relating to this e-mail. http://www.ml.com/email_terms/ -- For LINUX-390 subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390
Re: Any caveats moving root filesystem to LVM?
Some performance comparisons of LVM stiping are in chapter 8 of IBM's "Linux on IBM zSeries and S/390:Performance Measurement and Tuning" http://www.redbooks.ibm.com/redbooks/pdfs/sg246926.pdf -Original Message- From: Linux on 390 Port [mailto:[EMAIL PROTECTED] On Behalf Of Mrohs, Ray Sent: Monday, January 24, 2005 9:41 AM To: LINUX-390@VM.MARIST.EDU Subject: Re: Any caveats moving root filesystem to LVM? Striping does give you more paths into storage, but I haven't seen any performance studies of striped vs. non-striped LVM disks. If you have fast hardware and FICON, the advantage might not be that great. With striping, you lose the benefit of adding or removing physical volumes dynamically in the volume group (well, you still have to umount the file system briefly), which means you have to plan your file system growth really well, or take a chunk of down time to dump, resize, and reload whenever you need more space. We use the non-striped variety. Ray Mrohs Energy Information Administration U.S. Department of Energy -- For LINUX-390 subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390
Re: Any caveats moving root filesystem to LVM?
Striping does give you more paths into storage, but I haven't seen any performance studies of striped vs. non-striped LVM disks. If you have fast hardware and FICON, the advantage might not be that great. With striping, you lose the benefit of adding or removing physical volumes dynamically in the volume group (well, you still have to umount the file system briefly), which means you have to plan your file system growth really well, or take a chunk of down time to dump, resize, and reload whenever you need more space. We use the non-striped variety. Ray Mrohs Energy Information Administration U.S. Department of Energy -Original Message- From: David Andrews [mailto:[EMAIL PROTECTED] Sent: Monday, January 24, 2005 8:50 AM To: LINUX-390@VM.MARIST.EDU Subject: Re: Any caveats moving root filesystem to LVM? On Fri, 2005-01-21 at 16:02 -0500, Mrohs, Ray wrote: > My rule-of-thumb is to only use LVM when it's necessary, as in providing more > file system space than one minidisk can provide. There is also some striping value, no? (At least until Linux supports PAV.) -- David Andrews A. Duda and Sons, Inc. [EMAIL PROTECTED] -- For LINUX-390 subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390 -- For LINUX-390 subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390
Re: Any caveats moving root filesystem to LVM?
On Fri, 2005-01-21 at 16:02 -0500, Mrohs, Ray wrote: > My rule-of-thumb is to only use LVM when it's necessary, as in providing more > file system space than one minidisk can provide. There is also some striping value, no? (At least until Linux supports PAV.) -- David Andrews A. Duda and Sons, Inc. [EMAIL PROTECTED] -- For LINUX-390 subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390
Re: Any caveats moving root filesystem to LVM?
The one caveat I'd lay out is: DON'T. There are only a couple of arguments pro, and zillions of arguments con. The con argument that carries the most weight with me is that you can recover from most any other disk disaster _except_ the corruption of your root device. As has been suggested, I make all of the 1st level subdirectories separate mount points, and place their contents on non-root devices (LVMs, even!!). I only keep the directories needed for booting (/etc, /bin, /sbin, /boot, /lib ... CAUTION: this list might not be complete!! I'm working off the top of my head...) on the root device. Once you've done this, you won't have any reason to make your root device LVM -- it will turn out to be very manageable in terms of size. If you wanna use that extra space on the root device, go ahead and partition it, and give partition 2 over to LVM for use. Bottom line: there are a few disaster cases where having your root device LVMed would make your system unbootable. I haven't sat down to count them all, but they exist. I use LVM for everything except the root disk. LVM has lots of value for allocations that exceed physical device bounds: a well-administered root device isn't one of them. As to the quality of LVM overall: backup your filesystems with filesystem-independent tools (like tar, Amanda, TSM, etc). Do not trust disk surface in the long run: it _can_ go bad, LVM or not. LVM just adds one more layer of potential data scrambling ... never a good thing when you're up to your waist in a disaster. LVM works very well, and I've found it so far to be very reliable (2 years in service now); but you don't want to deal with that extra layer in a disaster context, and you gain nothing from putting a well-thought-out root device under LVM control. Peace, --Jim-- James S. Tison Senior Software Engineer TPF Laboratory / Architecture IBM Corporation Meum cerebrum nocet -- For LINUX-390 subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390 smime.p7s Description: S/MIME Cryptographic Signature
Re: Any caveats moving root filesystem to LVM?
__ Ranga Nathan / CSG Systems Programmer - Specialist; Technical Services; BAX Global Inc. Irvine-California Tel: 714-442-7591 Fax: 714-442-2840 "Post, Mark K" <[EMAIL PROTECTED]> Sent by: Linux on 390 Port 01/21/2005 04:17 PM Please respond to Linux on 390 Port To LINUX-390@VM.MARIST.EDU cc Subject Re: Any caveats moving root filesystem to LVM? Why would it be? We're talking about LVM versus non-LVM, not file systems. I got the impression that one should prefer ext2/ext3 for the root file system. I feel good now! Thanks Mark Post -Original Message- From: Linux on 390 Port [mailto:[EMAIL PROTECTED] On Behalf Of Ranga Nathan Sent: Friday, January 21, 2005 6:47 PM To: LINUX-390@VM.MARIST.EDU Subject: Re: Any caveats moving root filesystem to LVM? My root partition is reiserfs. So far I have not seen any problems. Is it a big NO NO? __ Ranga Nathan / CSG Systems Programmer - Specialist; Technical Services; BAX Global Inc. Irvine-California Tel: 714-442-7591 Fax: 714-442-2840 -- For LINUX-390 subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390 -- For LINUX-390 subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390
Re: Any caveats moving root filesystem to LVM?
Why would it be? We're talking about LVM versus non-LVM, not file systems. Mark Post -Original Message- From: Linux on 390 Port [mailto:[EMAIL PROTECTED] On Behalf Of Ranga Nathan Sent: Friday, January 21, 2005 6:47 PM To: LINUX-390@VM.MARIST.EDU Subject: Re: Any caveats moving root filesystem to LVM? My root partition is reiserfs. So far I have not seen any problems. Is it a big NO NO? __ Ranga Nathan / CSG Systems Programmer - Specialist; Technical Services; BAX Global Inc. Irvine-California Tel: 714-442-7591 Fax: 714-442-2840 -- For LINUX-390 subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390
Re: Any caveats moving root filesystem to LVM?
My root partition is reiserfs. So far I have not seen any problems. Is it a big NO NO? __ Ranga Nathan / CSG Systems Programmer - Specialist; Technical Services; BAX Global Inc. Irvine-California Tel: 714-442-7591 Fax: 714-442-2840 Adam Thornton <[EMAIL PROTECTED]> Sent by: Linux on 390 Port 01/21/2005 01:59 PM Please respond to Linux on 390 Port To LINUX-390@VM.MARIST.EDU cc Subject Re: Any caveats moving root filesystem to LVM? On Jan 21, 2005, at 3:40 PM, David Boyes wrote: > > 1) / as ext2 or ext3 (depending on distribution and platform -- usually > ext2 on zSeries) > This is the only place I disagree with David; I kind of recommend splitting /boot off and making that a small ext2 partition that contains the IPL record, initrd, and kernel. Then / can be LVM, or ext3 (you'd still need ext3 modules in your initrd of course, if it's not in the kernel), or whatever. This may be more hassle than it's worth in your environment. For our development machines, it certainly is. Adam -- For LINUX-390 subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390 -- For LINUX-390 subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390
Re: Any caveats moving root filesystem to LVM?
I used to take this approach also, until the umpteenth time I wanted to add just a little more space to a file system, and I had to wind up copying the entire contents off to a bigger piece of disk, verifying the copy is good, unmounting, remounting the new one, retiring the old piece of disk. Now, I just: umount /dev/vgname/lvname e2fsadm -L +??M /dev/vgname/lvname mount /dev/vgname/lvname I like that a lot better. Mark Post -Original Message- From: Linux on 390 Port [mailto:[EMAIL PROTECTED] On Behalf Of David Boyes Sent: Friday, January 21, 2005 4:41 PM To: LINUX-390@VM.MARIST.EDU Subject: Re: Any caveats moving root filesystem to LVM? -snip- 3) if a file system needs to be bigger than a physical volume, then use LVM and create ext3 filesystems on the logical volumes created by LVM. -- For LINUX-390 subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390
Re: Any caveats moving root filesystem to LVM?
On Jan 21, 2005, at 3:40 PM, David Boyes wrote: 1) / as ext2 or ext3 (depending on distribution and platform -- usually ext2 on zSeries) This is the only place I disagree with David; I kind of recommend splitting /boot off and making that a small ext2 partition that contains the IPL record, initrd, and kernel. Then / can be LVM, or ext3 (you'd still need ext3 modules in your initrd of course, if it's not in the kernel), or whatever. This may be more hassle than it's worth in your environment. For our development machines, it certainly is. Adam -- For LINUX-390 subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390
Re: Any caveats moving root filesystem to LVM?
> I forgot to mention in my original question that /boot wouldn't be in > LVM, it'd be an ordinary partition. On a separate minidisk, I hope. 8-) > Thank you for pointing out the fsck time consideration. Doesn't having > ext3 fs reduce the fsck time? Depends if you have ext3 built into the kernel. Also, do you need journaling on a file system that will take you at most a minute to fsck, and theoretically should have no significant number of files in it? It's a matter of minimizing the number of things that can go wrong at a time where there's not a lot of things that the OS can do to protect itself other than give up in a writhing heap of ashes. > My motivation in considering LVM is to get a general solution to my > Linux dasd constraints that I can use on all my Linux > guests. My Linux > guests start out as clones with about 4GB of disk; a default > install of > SLES 8. Different problem, then. What I use is: 1) / as ext2 or ext3 (depending on distribution and platform -- usually ext2 on zSeries) 2) all other filesystems as ext3 3) if a file system needs to be bigger than a physical volume, then use LVM and create ext3 filesystems on the logical volumes created by LVM. That works on pretty-much all flavors of Linux and all platforms, and if something goes horribly wrong, then diagnostics are pretty straightforward, and I can usually get the systems up to the point that you can at least try to fix things w/o a rescue system. -- For LINUX-390 subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390
Re: Any caveats moving root filesystem to LVM?
Yes, ext3 reduces/eliminates fsck time on a reboot. The thing to keep in mind is that the value of LVM is in being able to add/remove disk volumes from a volume group without having to move/rebuild the whole file system. If you break out all your file systems that don't have to be in the root file system and put them on LVM, you are left with very, very little space being used, and it should not change very often at all. Putting the root file system on LVM introduces unnecessary complexity for no real gain. See the note from Peter Abresch about the gymnastics he had to go through to correct a problem with an LVM volume group. (Which, by the way, he wrote up for me and I'm going to upload to the web site soon.) Not something I want to have to deal with to get my system up to the point where I can connect to it remotely. Mark Post -Original Message- From: Linux on 390 Port [mailto:[EMAIL PROTECTED] On Behalf Of Romanowski, John (OFT) Sent: Friday, January 21, 2005 4:18 PM To: LINUX-390@VM.MARIST.EDU Subject: Re: Any caveats moving root filesystem to LVM? I forgot to mention in my original question that /boot wouldn't be in LVM, it'd be an ordinary partition. Thank you for pointing out the fsck time consideration. Doesn't having ext3 fs reduce the fsck time? My motivation in considering LVM is to get a general solution to my Linux dasd constraints that I can use on all my Linux guests. My Linux guests start out as clones with about 4GB of disk; a default install of SLES 8. -- For LINUX-390 subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390
Re: Any caveats moving root filesystem to LVM?
I forgot to mention in my original question that /boot wouldn't be in LVM, it'd be an ordinary partition. Thank you for pointing out the fsck time consideration. Doesn't having ext3 fs reduce the fsck time? My motivation in considering LVM is to get a general solution to my Linux dasd constraints that I can use on all my Linux guests. My Linux guests start out as clones with about 4GB of disk; a default install of SLES 8. -Original Message- From: Linux on 390 Port [mailto:[EMAIL PROTECTED] On Behalf Of David Boyes Sent: Friday, January 21, 2005 3:38 PM To: LINUX-390@VM.MARIST.EDU Subject: Re: Any caveats moving root filesystem to LVM? > But if LVM is so hard to fix then why use LVM even for /home, > /opt, etc? Because LVM is necessary to allow those file systems to be larger than a single physical volume, and those filesystems are not usually critical during early stages of the boot process where things are still a little fragile. LVM != RAID, although they share some common ideas and techniques. For most systems, / and /boot are never that large (so fsck time is negligible), and *are* critical during boot -- general KISS principle applies in known critical situations. -- For LINUX-390 subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390 -- For LINUX-390 subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390
Re: Any caveats moving root filesystem to LVM?
Thank you; it sounds like the identical LVM volume group names do matter for rescue considerations, but in your situation you can get around the name-collision. -Original Message- From: Linux on 390 Port [mailto:[EMAIL PROTECTED] On Behalf Of Mrohs, Ray Sent: Friday, January 21, 2005 4:03 PM To: LINUX-390@VM.MARIST.EDU Subject: Re: Any caveats moving root filesystem to LVM? My rule-of-thumb is to only use LVM when it's necessary, as in providing more file system space than one minidisk can provide. I put /temp, /var, etc. on their own minidisks so that errant processes cannot accidentally fill all the available free space and crash the system. Our LVMs hold databases, so if we need to repair one from another Linux, its just a matter of unmounting the existing LVM, and then linking to the disks that need repairs, in which case identical volume and group names don't matter. Yes simple is much better, especially at 3AM! Ray Mrohs Energy Information Administration U.S. Department of Energy -Original Message- From: Romanowski, John (OFT) [mailto:[EMAIL PROTECTED] Sent: Friday, January 21, 2005 12:05 PM To: LINUX-390@VM.MARIST.EDU Subject: Any caveats moving root filesystem to LVM? Anyone have experiences moving root file system to Logical Volume Manager (LVM)? We run SLES8 under VM on S390 with Linux guests cloned from a 2-dasd (3390mod3's) linux image with "/" on one pack and /usr on the other pack. I'd like more flexibility to use the free disk space from each pack as a global pool of free space so I'm evaluating converting the existing Linux file systems to LVM with root file system in LVM. I'm comfortable on how to create the logical volumes and copy the existing Linux file systems into it, but I'm not clear on how the LVM-based Linux will boot up and how I'd "rescue" such systems later if needed. 1) At IPL how will the Suse initrd find the Volume group and mount the root file system? I'm going to change /boot/zipl/parmfile to say "root=/dev/vg/v1" to request the logical volume be mounted as root fs; my /etc/sysconfig/kernel will be coded to load lvm-mod via INITRD_MODULES="jbd ext3 dasd_diag_mod lvm-mod"; and I'll run mkinitrd and zipl. But is that enough to get a logical volume mounted as the root file fs by initrd? 2) After converting to LVM I'll have a bunch of Linux guests whose Volume Group name and logical volume names are identical (as if I'd originally cloned them from an LVM-based Linux image). For rescue purposes, can one of these Linux's CP LINK to and mount another of these Linux's volume groups and volumes given it's already running with the same volume group & volume names itself? Any tips on this are appreciated. -- For LINUX-390 subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390 -- For LINUX-390 subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390 -- For LINUX-390 subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390
Re: Any caveats moving root filesystem to LVM?
My rule-of-thumb is to only use LVM when it's necessary, as in providing more file system space than one minidisk can provide. I put /temp, /var, etc. on their own minidisks so that errant processes cannot accidentally fill all the available free space and crash the system. Our LVMs hold databases, so if we need to repair one from another Linux, its just a matter of unmounting the existing LVM, and then linking to the disks that need repairs, in which case identical volume and group names don't matter. Yes simple is much better, especially at 3AM! Ray Mrohs Energy Information Administration U.S. Department of Energy -Original Message- From: Romanowski, John (OFT) [mailto:[EMAIL PROTECTED] Sent: Friday, January 21, 2005 12:05 PM To: LINUX-390@VM.MARIST.EDU Subject: Any caveats moving root filesystem to LVM? Anyone have experiences moving root file system to Logical Volume Manager (LVM)? We run SLES8 under VM on S390 with Linux guests cloned from a 2-dasd (3390mod3's) linux image with "/" on one pack and /usr on the other pack. I'd like more flexibility to use the free disk space from each pack as a global pool of free space so I'm evaluating converting the existing Linux file systems to LVM with root file system in LVM. I'm comfortable on how to create the logical volumes and copy the existing Linux file systems into it, but I'm not clear on how the LVM-based Linux will boot up and how I'd "rescue" such systems later if needed. 1) At IPL how will the Suse initrd find the Volume group and mount the root file system? I'm going to change /boot/zipl/parmfile to say "root=/dev/vg/v1" to request the logical volume be mounted as root fs; my /etc/sysconfig/kernel will be coded to load lvm-mod via INITRD_MODULES="jbd ext3 dasd_diag_mod lvm-mod"; and I'll run mkinitrd and zipl. But is that enough to get a logical volume mounted as the root file fs by initrd? 2) After converting to LVM I'll have a bunch of Linux guests whose Volume Group name and logical volume names are identical (as if I'd originally cloned them from an LVM-based Linux image). For rescue purposes, can one of these Linux's CP LINK to and mount another of these Linux's volume groups and volumes given it's already running with the same volume group & volume names itself? Any tips on this are appreciated. -- For LINUX-390 subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390 -- For LINUX-390 subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390
Re: Any caveats moving root filesystem to LVM?
> But if LVM is so hard to fix then why use LVM even for /home, > /opt, etc? Because LVM is necessary to allow those file systems to be larger than a single physical volume, and those filesystems are not usually critical during early stages of the boot process where things are still a little fragile. LVM != RAID, although they share some common ideas and techniques. For most systems, / and /boot are never that large (so fsck time is negligible), and *are* critical during boot -- general KISS principle applies in known critical situations. -- For LINUX-390 subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390
Re: Any caveats moving root filesystem to LVM?
But if LVM is so hard to fix then why use LVM even for /home, /opt, etc? If LVM is fixable from a running Linux I have a lot of running Linux guests or can use the Suse installation system booted from the guest's RDR, no? What are the specific recovery issues? -Original Message- From: Linux on 390 Port [mailto:[EMAIL PROTECTED] On Behalf Of Post, Mark K Sent: Friday, January 21, 2005 12:39 PM To: LINUX-390@VM.MARIST.EDU Subject: Re: Any caveats moving root filesystem to LVM? Don't do it. Keep your root file system as plain vanilla ext3, and move other things to LVM (using ext3 as well): /home /opt /tmp /usr /var You really, really, don't want to have to fix LVM to get your system up an running if it ever comes to that. Mark Post -Original Message- From: Linux on 390 Port [mailto:[EMAIL PROTECTED] On Behalf Of Romanowski, John (OFT) Sent: Friday, January 21, 2005 12:05 PM To: LINUX-390@VM.MARIST.EDU Subject: Any caveats moving root filesystem to LVM? Anyone have experiences moving root file system to Logical Volume Manager (LVM)? We run SLES8 under VM on S390 with Linux guests cloned from a 2-dasd (3390mod3's) linux image with "/" on one pack and /usr on the other pack. I'd like more flexibility to use the free disk space from each pack as a global pool of free space so I'm evaluating converting the existing Linux file systems to LVM with root file system in LVM. I'm comfortable on how to create the logical volumes and copy the existing Linux file systems into it, but I'm not clear on how the LVM-based Linux will boot up and how I'd "rescue" such systems later if needed. -snip- -- For LINUX-390 subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390 -- For LINUX-390 subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390
Re: Any caveats moving root filesystem to LVM?
Tom, Yes. You need at least one partition where you are _sure_ that the data written will be going where it needs to be for the IPL process to find it. Mark Post -Original Message- From: Linux on 390 Port [mailto:[EMAIL PROTECTED] On Behalf Of Tom Duerbusch Sent: Friday, January 21, 2005 12:40 PM To: LINUX-390@VM.MARIST.EDU Subject: Re: Any caveats moving root filesystem to LVM? Is it true, note that I haven't tried it, that if you try to use LVM for everything, you still need a /boot volume, just a few cylinders, for IPL purposes? I'm thinking that LVM is a software raid, so you have to have the software running before you can use it. -- For LINUX-390 subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390
Re: Any caveats moving root filesystem to LVM?
I have done this on a fresh install. /boot was it's own filesystem / was in lvm vgwhatever SUSE appends LVM to the /etc/sysconfig/modules so that the lvm module is available in the initial ramd. HTH Wh Michael MacIsaac <[EMAIL PROTECTED]> wrote: > At IPL how will the Suse initrd find the Volume group and mount the root file system? I don't think it can. We boot from a 100 cyl /boot fs and have a logical volume mounted over /. Maybe the easiest way is to try this type of migration is to play tricks with minidisks: - Shut down the Linux to be migrated - Define a new user ID with the two or three new volumes as well as the volumes with the old / and /usr/ as different minidisks - Install a fresh sles8 with a 100 cyl /boot and a two or three volume LVM over / - Temporarily mount and recursively copy the old / to the new / - Do the same for /usr/ - Reboot and cross your fingers - if it doesn't work, at least the old user ID is unchanged and can still be booted I've never tried this, but it sounds good on "paper". (During the time I was writing this I see Mark and Brandon's posts. I tend to agree - you may want to avoid logical volumes over / altogether.) "Mike MacIsaac" (845) 433-7061 -- For LINUX-390 subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390 -- For LINUX-390 subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390
Re: Any caveats moving root filesystem to LVM?
> At IPL how will the Suse initrd find the Volume group and mount the root file system? I don't think it can. We boot from a 100 cyl /boot fs and have a logical volume mounted over /. Maybe the easiest way is to try this type of migration is to play tricks with minidisks: - Shut down the Linux to be migrated - Define a new user ID with the two or three new volumes as well as the volumes with the old / and /usr/ as different minidisks - Install a fresh sles8 with a 100 cyl /boot and a two or three volume LVM over / - Temporarily mount and recursively copy the old / to the new / - Do the same for /usr/ - Reboot and cross your fingers - if it doesn't work, at least the old user ID is unchanged and can still be booted I've never tried this, but it sounds good on "paper". (During the time I was writing this I see Mark and Brandon's posts. I tend to agree - you may want to avoid logical volumes over / altogether.) "Mike MacIsaac" <[EMAIL PROTECTED]> (845) 433-7061 -- For LINUX-390 subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390
Re: Any caveats moving root filesystem to LVM?
Post, Mark K wrote: Don't do it. Keep your root file system as plain vanilla ext3, and move other things to LVM (using ext3 as well): /home /opt /tmp /usr /var You really, really, don't want to have to fix LVM to get your system up an running if it ever comes to that. Mark Post Agreed. When we have to recover HP-UX systems with an LVM'd root disk it really is a pain. *Brandon -- For LINUX-390 subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390
Re: Any caveats moving root filesystem to LVM?
Is it true, note that I haven't tried it, that if you try to use LVM for everything, you still need a /boot volume, just a few cylinders, for IPL purposes? I'm thinking that LVM is a software raid, so you have to have the software running before you can use it. Tom Duerbusch THD Consulting >>> [EMAIL PROTECTED] 01/21/05 11:12 AM >>> I had problems moving root file system to LVM. It has to be done with a lot of care. However, if you move /home /var /local /opt etc to LVM, then there is much less reason for moving root fs to LVM. That is what I did. We run SLES9. I used Yast to do all LVM work, eventhough I knew how to do it from command line. I then did mkinitrd and zipl. Yast updates /etc/fstab when creating LVM. However later when you actually move and create new mount points, you need to make sure that /etc/fstab is correct. This was one source of problem for me. Then I re-booted the guest. Having another guest to mount sick paritions was a great. __ Ranga Nathan / CSG Systems Programmer - Specialist; Technical Services; BAX Global Inc. Irvine-California Tel: 714-442-7591 Fax: 714-442-2840 "Romanowski, John (OFT)" <[EMAIL PROTECTED]> Sent by: Linux on 390 Port 01/21/2005 09:04 AM Please respond to Linux on 390 Port To LINUX-390@VM.MARIST.EDU cc Subject Any caveats moving root filesystem to LVM? Anyone have experiences moving root file system to Logical Volume Manager (LVM)? We run SLES8 under VM on S390 with Linux guests cloned from a 2-dasd (3390mod3's) linux image with "/" on one pack and /usr on the other pack. I'd like more flexibility to use the free disk space from each pack as a global pool of free space so I'm evaluating converting the existing Linux file systems to LVM with root file system in LVM. I'm comfortable on how to create the logical volumes and copy the existing Linux file systems into it, but I'm not clear on how the LVM-based Linux will boot up and how I'd "rescue" such systems later if needed. 1) At IPL how will the Suse initrd find the Volume group and mount the root file system? I'm going to change /boot/zipl/parmfile to say "root=/dev/vg/v1" to request the logical volume be mounted as root fs; my /etc/sysconfig/kernel will be coded to load lvm-mod via INITRD_MODULES="jbd ext3 dasd_diag_mod lvm-mod"; and I'll run mkinitrd and zipl. But is that enough to get a logical volume mounted as the root file fs by initrd? 2) After converting to LVM I'll have a bunch of Linux guests whose Volume Group name and logical volume names are identical (as if I'd originally cloned them from an LVM-based Linux image). For rescue purposes, can one of these Linux's CP LINK to and mount another of these Linux's volume groups and volumes given it's already running with the same volume group & volume names itself? Any tips on this are appreciated. -- For LINUX-390 subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390 -- For LINUX-390 subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390 -- For LINUX-390 subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390
Re: Any caveats moving root filesystem to LVM?
Don't do it. Keep your root file system as plain vanilla ext3, and move other things to LVM (using ext3 as well): /home /opt /tmp /usr /var You really, really, don't want to have to fix LVM to get your system up an running if it ever comes to that. Mark Post -Original Message- From: Linux on 390 Port [mailto:[EMAIL PROTECTED] On Behalf Of Romanowski, John (OFT) Sent: Friday, January 21, 2005 12:05 PM To: LINUX-390@VM.MARIST.EDU Subject: Any caveats moving root filesystem to LVM? Anyone have experiences moving root file system to Logical Volume Manager (LVM)? We run SLES8 under VM on S390 with Linux guests cloned from a 2-dasd (3390mod3's) linux image with "/" on one pack and /usr on the other pack. I'd like more flexibility to use the free disk space from each pack as a global pool of free space so I'm evaluating converting the existing Linux file systems to LVM with root file system in LVM. I'm comfortable on how to create the logical volumes and copy the existing Linux file systems into it, but I'm not clear on how the LVM-based Linux will boot up and how I'd "rescue" such systems later if needed. -snip- -- For LINUX-390 subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390
Re: Any caveats moving root filesystem to LVM?
On Fri, 2005-01-21 at 11:12, Ranga Nathan wrote: > I had problems moving root file system to LVM. It has to be done with a > lot of care. However, if you move /home /var /local /opt etc to LVM, then > there is much less reason for moving root fs to LVM. That is what I did. > We run SLES9. I used Yast to do all LVM work, eventhough I knew how to do > it from command line. I then did mkinitrd and zipl. > We do the same thing, except under SLES8. Our root filesystems only have 100MB (with 23MB used). Our reason is, we just like one less layer to repair in case of a corruption. I'll have to try Yast2 the next time there's LVM work. -- For LINUX-390 subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390
Re: Any caveats moving root filesystem to LVM?
I had problems moving root file system to LVM. It has to be done with a lot of care. However, if you move /home /var /local /opt etc to LVM, then there is much less reason for moving root fs to LVM. That is what I did. We run SLES9. I used Yast to do all LVM work, eventhough I knew how to do it from command line. I then did mkinitrd and zipl. Yast updates /etc/fstab when creating LVM. However later when you actually move and create new mount points, you need to make sure that /etc/fstab is correct. This was one source of problem for me. Then I re-booted the guest. Having another guest to mount sick paritions was a great. __ Ranga Nathan / CSG Systems Programmer - Specialist; Technical Services; BAX Global Inc. Irvine-California Tel: 714-442-7591 Fax: 714-442-2840 "Romanowski, John (OFT)" <[EMAIL PROTECTED]> Sent by: Linux on 390 Port 01/21/2005 09:04 AM Please respond to Linux on 390 Port To LINUX-390@VM.MARIST.EDU cc Subject Any caveats moving root filesystem to LVM? Anyone have experiences moving root file system to Logical Volume Manager (LVM)? We run SLES8 under VM on S390 with Linux guests cloned from a 2-dasd (3390mod3's) linux image with "/" on one pack and /usr on the other pack. I'd like more flexibility to use the free disk space from each pack as a global pool of free space so I'm evaluating converting the existing Linux file systems to LVM with root file system in LVM. I'm comfortable on how to create the logical volumes and copy the existing Linux file systems into it, but I'm not clear on how the LVM-based Linux will boot up and how I'd "rescue" such systems later if needed. 1) At IPL how will the Suse initrd find the Volume group and mount the root file system? I'm going to change /boot/zipl/parmfile to say "root=/dev/vg/v1" to request the logical volume be mounted as root fs; my /etc/sysconfig/kernel will be coded to load lvm-mod via INITRD_MODULES="jbd ext3 dasd_diag_mod lvm-mod"; and I'll run mkinitrd and zipl. But is that enough to get a logical volume mounted as the root file fs by initrd? 2) After converting to LVM I'll have a bunch of Linux guests whose Volume Group name and logical volume names are identical (as if I'd originally cloned them from an LVM-based Linux image). For rescue purposes, can one of these Linux's CP LINK to and mount another of these Linux's volume groups and volumes given it's already running with the same volume group & volume names itself? Any tips on this are appreciated. -- For LINUX-390 subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390 -- For LINUX-390 subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390