Take the 2-minute tour ×
Server Fault is a question and answer site for system and network administrators. It's 100% free, no registration required.

VMware allows to extend the size of a virtual disk online - when the VM is running. The next expected steps for Linux system are:

  1. extend the partition: delete and create a larger one with fdisk
  2. extend the PV size with pvresize
  3. use free extents for lvresize operations
  4. and then resize2fs for file system

But I am stuck on the first step: fdisk and sfdisk still display the old size for the disk.

My disk is a SCSI virtual disk connected thanks to the virtual LSI Logic controller.

How to refresh the virtual disk size and partition table information available in Linux kernel without reboot ?

As far as I know all that steps are possible for a running Windows, without reboot and even without any user actions thanks to VMWare tools. On Linux, I expects to do all steps online too and I already know steps 2, 3 and 4 work online. But the first one - change partition size declared in the partition table (still) seems to require a reboot.

Update: My system is a Debian Lenny with kernel 2.6.26 and the disk I have extended is the main disk with a large PV containing the "root" LV for "/".

share|improve this question

5 Answers 5

up vote 5 down vote accepted

You can do this without a reboot. pvresize doesn't resize the physical volume until the partition is updated with the added space. You must fdisk the partition and recreate it with the the new full size of the disk, after you can rescan'd the drives for it to see the extra space in the first place.

More info: http://theducks.org/2009/11/expanding-lvm-partitions-in-vmware-on-the-fly/

share|improve this answer
    
I agree but whatever I did to rescan the SCSI drive, I never got the new disk size visible in fdisk, so I was stuck until a first reboot. And "partprobe" is a new command for me. Definitely I have to test this procedure... Thank you. –  Yves Martin Oct 9 '12 at 19:39
    
For information partprobe is a command from parted project. –  Yves Martin Mar 25 '13 at 15:32

You need to rescan the disks before you can make the bigger partition.

In Centos you can do this by

ls /sys/class/scsi_host

then for each host

echo "- - -" > /sys/class/scsi_host/host#/scan

(replace # with the number)

There is also one more step to the above which is expanding ext or whatever filesystem you are using once you have resized the partition.

You are still going to have to unmount that partition though at some point. What we tend to do is add a 2nd vmware disk and use lvm to extend onto the new disk (and reduce off the old if it is a replacement) as this allows the whole process to happen live.

share|improve this answer
    
My system is a Debian and I run rescan-scsi-bus.sh from scsitools. I think it is equivalent to the command you propose on /sys/class/scsi_host but "fdisk" still display the old size on my disk... –  Yves Martin Apr 11 '12 at 7:47
    
Does it recognise a new vmware disk when you rescan? –  JamesRyan Apr 11 '12 at 10:16
    
No. This disk exists already, it is not new. The disk size has been increased in vSphere client. –  Yves Martin Apr 11 '12 at 12:23
    
I realise that but you should try adding a new one to see if your scan picks it up to narrow down the problem –  JamesRyan Apr 12 '12 at 9:27
    
I agree with you and I already know by practice that add a new SCSI disk to a Linux machine can be done online for all the steps listed. As the VM may run for 10 years... I fear the total number of disks in a near future... –  Yves Martin Apr 12 '12 at 9:50

You've not provided us enough detail to tell you the exact commands you'll need but essentially you'll need to use the lvextend command to extend the logical volume, then the e2fsck command and then the resize2fs command to actually expand your filesystem. Each of these commands will need additional parameters, specifically device and filesystem information, that we can't provide but you'll need to know these, just use the --help option for each command to tell you how to use them specifically, plus you'll probably end up using the pvdisplay, lvdisplay and mount commands to help fill out these parameters.

share|improve this answer
    
I know all the commands for other steps. My question is: how to grow the primary partition used as PV without reboot ? –  Yves Martin Apr 10 '12 at 20:36
    
Sorry, do you just mean pvresize? –  Chopper3 Apr 10 '12 at 20:39
    
I agree my question was not explicit enough (even with the bold part). pvresize is mentioned as step 2. I am stuck at step 1: increase the "PCAT" partition size. –  Yves Martin Apr 10 '12 at 20:45
    
Have you increased the .vmdk size via the VSClient yet? –  Chopper3 Apr 10 '12 at 21:18
1  
To confirm /sbin/rescan-scsi-bus dooes not change existing disk size. But echo "1" > '/sys/class/scsi_disk/0:0:1:0/device/rescan' does –  Yves Martin Jul 21 at 16:13

As far as the root file system / is mounted on the disk that has been resized, the partition table and disk size are not refreshed by a SCSI rescan with Linux 2.6.26.

I really hope it will be better soon with newer kernel versions.

So I had to:

  • reboot a first time to see the new disk size in fdisk
  • delete the old primary PV partion in fdisk
  • create a partition entry with the same number and start sector until the end of disk
  • reboot a second time OR run partprobe from parted package only if / is not mounted there
  • Now I can run pvresize to get new free space, lvextend and resize2fs to allocate some more space to a file system

I have been recommended to simply discard that stupid old partition table and run pvcreate directly on the device as Grub2 is able to load a kernel image directly from a file system on a LVM partition.

share|improve this answer

The other answered provided do not address your question, I've identified the correct command to rescan an already connected disk.

We must rescan your already connected disk, first identify which disk you want to rescan.

ls /sys/class/scsi_disk/

In my example, I see a symlink named 0:0:0:0, so we rescan this scsi-disk.

echo '1' > /sys/class/scsi_disk/0\:0\:0\:0/device/rescan

I just extended my VMware disk also, and had to scour other answers to find the correct command. Hopefully this will save future searchers from futile attempts.

share|improve this answer

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.