Cloning Solaris & Linux servers

[worzelGummidge]

The easiest solution is to use NetBackup because it will be the same solution for both and can be recovered from bare metal.

After a full backup of everything for each server, using the same hardware configuration, Netbackup would restore a base image to each machine and then perform a full restore.


Without NetBackup, to just restore the server into a state that it was in you will be using a Solaris Jet server along with a flar image (big zip file). Solaris 8 would be a problem as integrating this with flar and Jet can get a bit tricky for the disk mirroring.

For the Linux server then use Kickstart to build the back OS. No idea what the equivalent for flar is on Linux though.

NetBackup would be the corporate solution that would always work.

The Solaris solutions would work and restore to a point when you took the flar image. Tricky for Solaris 8 though.

Linux - Not too sure.

[stek]

Nice.

I thought that it should also be possible to use a PIXE boot setup of some sort, can't remember what's possible with SPARC. Not used one rwally since Solaris 6.
Thinking again, isn't PIXE just like JumpStart? I.e. you can boot and install a bare metal box from the eeprom?

SPARC (and POWER) traditionally uses BOOTP but can WAN boot too (later ones) basically over http(s) but BOOTP'ing is only a small part of what Jumpstart does, it builds the full box, i.e. configures filesystem slices, setups NIC, installs additional packages, patches, build zones/containers etc.

[darrylmg]

The V440 is POS, but the T2000 is fully partition-able, like IBM LPAR's etc, proper firmware hypervisor, so it can offer virtualisation from the firmware, not from another layered OS like VMware, KVM etc...

T2000 pretty old now, T1 processor, but with 32gb RAM it can virtualise 10 or so full SPARC LDOMS and do it well. Added advantage is that if using file-based disk images, the DR is just a copy of a file. Sorted. T2000's cheap as chips now, 100 quid or so off Ebay so a DR spare is no issue.

Nice.

I thought that it should also be possible to use a PIXE boot setup of some sort, can't remember what's possible with SPARC. Not used one rwally since Solaris 6.
Thinking again, isn't PIXE just like JumpStart? I.e. you can boot and install a bare metal box from the eeprom?

[stek]

The V440 is POS, but the T2000 is fully partition-able, like IBM LPAR's etc, proper firmware hypervisor, so it can offer virtualisation from the firmware, not from another layered OS like VMware, KVM etc...

T2000 pretty old now, T1 processor, but with 32gb RAM it can virtualise 10 or so full SPARC LDOMS and do it well. Added advantage is that if using file-based disk images, the DR is just a copy of a file. Sorted. T2000's cheap as chips now, 100 quid or so off Ebay so a DR spare is no issue.

[Cliphead]

Thanks all for the advice. No decision has been made yet on where they want to go with this but I'll need at least to propose a solution should it be required.

The biggest problem is not having all the information I need, getting answers is like drawing teeth out of this mob.

[unixman]

You're looking for a bare metal recovery plan, which is not a trivial thing to implement. It will be a bit more than just copying a disk or similar. The procedure will depend on the setup, the time window they allow for recovery, the type of storage, the type of backups and other things.

I'm slightly confused on why they need a bare metal recover when the system is already clustered, but no matter. It does sound like a good opportunity to obtain more work.

[stek]

Sunfire T2000 and V440's so not sure, only got hit with this today.

Would something like clonezilla do the job or revert to rsync after building the partitions for the Linux boxes?

Pretty ancient Sparcs. I'd use a combination of flarchive for the data and Jumpstart to boot from.

Or mirror the disks, split the mirrors and remove the disks and keep safe!

Unfortunately neither Solaris nor Linux have anything like bootable mksysb in AIX and HPUX designed to recover from bare metal.

[Cliphead]

Another option could be to set up a DR server with the same setup, and just use rsync to make any changes on live go to DR

Do you actually need to clone the whole servers state ?

I need the whole server state for the remote possibilty that the worst case happens. It may be just a paper exercise but I suspect it may have to be demonstrated for real.

[MicrosoftBob]

Another option could be to set up a DR server with the same setup, and just use rsync to make any changes on live go to DR

Do you actually need to clone the whole servers state ?

[Cliphead]

Are the Solaris nodes SPARC?

Sunfire T2000 and V440's so not sure, only got hit with this today.

Would something like clonezilla do the job or revert to rsync after building the partitions for the Linux boxes?

[stek]

Flar on Solaris, or even ufsdump if not on ZFS, neither of these quick to restore from bare. Same for whatever RHEL/centos offers (probably nowt).

What you really need for bare-metal is AI/Jumpstart for Solaris and Kickstart for RHEL/Centos. Even better is xCAT, which will do both. Lot of configuring to do though not sure it's worth it for a few nodes.




Are the Solaris nodes SPARC?

[Cliphead]

Anybody done this? Any advice?

Two node high availability cluster with raid 5. Need to clone or backup / image one of the nodes for a fast restore in case of a hardware failure.

Solaris 8 & 10, CentOS 5.x and 6.x