I guess it is very common for administrative purpose, to dump and restore a CentOS 7 system. I usually use dump/restore commands. However, I
11 thoughts on - How To Dump/restore A CentOS 7 System
Though I can not answer OP’s question, I have question of my own.
Is this really routine (often) task for Linux sysadmins? I used something like that to replicate cluster nodes in the past, but kickstart would be routine task for me. dump/restore sounds like routine from MS Windows world (I hear they “re-image” system if something goes wrong ;-)
Am I wrong? Do we in Linux world do this routinely?
Valeri
You are not wrong. However, I will point out first that dump and restore are utilities that have been around the Unix/Linux world for a very long time, rather than something from the M$ world.
The issue of how to restore/copy a system installation is open to discussion these days. I have recently been in a situation where duplicating identical machines is done conveniently with dump and restore. I have also been in situations where installing or reinstalling a system of slightly different configuration is easily accomplished via kickstart.
It mostly depends on the situation to be addressed at the moment, and the tools available. For instance, to perform a one-time installation when you do not have kickstart set up on your network is a significant amount of work, and may not be worth the effort of kickstart set-up.
This is one of the benefits of decades of development. More tools are available to handle the installation requirements.
“God, root, what is difference ?” Piter from UserFriendly
I would not say routinely, but I would say crucially.
The poster child for dump/restore is a machine with commercial software that is difficult to install or customize, especially one with an RDBMS system large enough to make dumping and restoring the data tables an onerous task.
The usual workflow — kickstart and puppet/ansible/etc — doesn’t work in that situation.
I have had good luck with Clonezilla for both Linux and Windows machines.
It boots from CD and can put the image of the system in an assortment of places. I personally, dump to one of the large external USB drives.
Restoring to a new disk works fine so long as the disk is of equal size or larger than the original.
Valeri Galtsev wrote:
I have never used dump/restore.
Someone mentioned commercial software – I’ve cloned systems, esp. compute nodes in a cluster – with rsync.
mark
Problem with rsync clones, they are asynchronous each file is copied separately so if the system is live and making changes, for instance a database server, the copy is not coherent.
the ZFS file system is perfect for this stuff, you can snap shot and then ZFS send the snapshot or a delta between that snapshot and a previous snapshot and that creates a totally coherent copy made at a single point in time atomically.
Le 25/09/2019
One of the advantages of UEFI — The bootloader is just a fat32
partition with a couple of UEFI executables and the grub.cfg.
—
Jonathan Billings
–This hasn’t been addressed yet. After restoring the partition images from tape (presumably using a rescue CD), how does one make the system bootable?
BTW, I just saw an article on tape backup and how the media are still incredibly cheaper than disks (and far cheaper than SSD), but the drives are extremely expensive (like $3k) so prohibitive for single-server installations like the home user.
11 thoughts on - How To Dump/restore A CentOS 7 System
Though I can not answer OP’s question, I have question of my own.
Is this really routine (often) task for Linux sysadmins? I used something like that to replicate cluster nodes in the past, but kickstart would be routine task for me. dump/restore sounds like routine from MS Windows world (I hear they “re-image” system if something goes wrong ;-)
Am I wrong? Do we in Linux world do this routinely?
Valeri
You are not wrong. However, I will point out first that dump and restore are utilities that have been around the Unix/Linux world for a very long time, rather than something from the M$ world.
The issue of how to restore/copy a system installation is open to discussion these days. I have recently been in a situation where duplicating identical machines is done conveniently with dump and restore. I have also been in situations where installing or reinstalling a system of slightly different configuration is easily accomplished via kickstart.
It mostly depends on the situation to be addressed at the moment, and the tools available. For instance, to perform a one-time installation when you do not have kickstart set up on your network is a significant amount of work, and may not be worth the effort of kickstart set-up.
This is one of the benefits of decades of development. More tools are available to handle the installation requirements.
—
Ron Loftin reloftin@twcny.rr.com
“God, root, what is difference ?” Piter from UserFriendly
I would not say routinely, but I would say crucially.
The poster child for dump/restore is a machine with commercial software that is difficult to install or customize, especially one with an RDBMS system large enough to make dumping and restoring the data tables an onerous task.
The usual workflow — kickstart and puppet/ansible/etc — doesn’t work in that situation.
I have had good luck with Clonezilla for both Linux and Windows machines.
It boots from CD and can put the image of the system in an assortment of places. I personally, dump to one of the large external USB drives.
Restoring to a new disk works fine so long as the disk is of equal size or larger than the original.
Valeri Galtsev wrote:
I have never used dump/restore.
Someone mentioned commercial software – I’ve cloned systems, esp. compute nodes in a cluster – with rsync.
mark
Problem with rsync clones, they are asynchronous each file is copied separately so if the system is live and making changes, for instance a database server, the copy is not coherent.
the ZFS file system is perfect for this stuff, you can snap shot and then ZFS send the snapshot or a delta between that snapshot and a previous snapshot and that creates a totally coherent copy made at a single point in time atomically.
Le 25/09/2019
One of the advantages of UEFI — The bootloader is just a fat32
partition with a couple of UEFI executables and the grub.cfg.
—
Jonathan Billings
–This hasn’t been addressed yet. After restoring the partition images from tape (presumably using a rescue CD), how does one make the system bootable?
BTW, I just saw an article on tape backup and how the media are still incredibly cheaper than disks (and far cheaper than SSD), but the drives are extremely expensive (like $3k) so prohibitive for single-server installations like the home user.
<https://www.techradar.com/news/theres-one-crucial-way-tape-still-trounces-ssds-and-hard-drives-when-it-comes-to-storage>
–If you use LVM (CentOS default), you can use its snapshot feature.
I am using mondorescue (mondoarchive, mondorestore) for years with success: http://www.mondorescue.org/
Allows bare metal recovery. I have tested and it works fine in CentOS
5,6,7.
In CentOS 7, after restore, I have had some issues, but I managed to resolve them. (Details in mailing list archives.)
Cheers, Nick