Google Groups no longer supports new Usenet posts or subscriptions. Historical content remains viewable.
Dismiss

Backing up Solaris 10 servers

59 views
Skip to first unread message

BertieB...@gmail.com

unread,
Nov 6, 2012, 11:51:36 AM11/6/12
to
Got 5-6 Solaris 10 servers in remote location - power work being done.

At the moment, we just back up the application database that these servers run with the idea that if it all goes wrong we can rebuild and then restore the application.
However, requirement is to provide a complete backup in the event of issues caused by possible power failure.

Previous experience has shown that unexpected results can occur with sudden power loss on solaris servers - we've probably all experienced that!

Any suggestions for ideas? I was thinking:-

1. Flash images (ideally it would be nice to set up a jumpstart server)
2. ufsdump
3. dd to spare disks (some of the servers have 4 disks installed with just one disk in use).

Also, got one HPUX server and a load of Windows servers to do as well

Cydrome Leader

unread,
Nov 6, 2012, 12:05:33 PM11/6/12
to
4. netbackup

BertieB...@gmail.com

unread,
Nov 6, 2012, 1:10:49 PM11/6/12
to
Take it you mean Symantec Netbackup here?

Assuming this needs a server somewhere and then the installation of clients on each client server?

Richard B. Gilbert

unread,
Nov 6, 2012, 1:17:25 PM11/6/12
to
How you make your backups is less important than that you DO MAKE
BACKUPS!!!!!! If you have the means, make two copies and send one off site.

You also need to verify your ability to restore that backup! A backup
that you cannot restore is worthless. Ideally, you should be able to
restore your backup to your primary site and/or an emergency site.

Ian Collins

unread,
Nov 6, 2012, 1:19:07 PM11/6/12
to
4. Create a pool on spare disk(s) and zfs send your data there. If you
have space, rsync the data from the other OS systems there as well.

5. Commercial backup software!

--
Ian Collins

Cydrome Leader

unread,
Nov 6, 2012, 1:20:41 PM11/6/12
to
yes.

it can backup windows and unix. It will require a server with the tape
drive(s) and then client software on each client.

If you completely destroy a solaris machine you can reinstall from a flar,
then update the files by doing a restore with netbackup.

there's not really any type of bare metal reinstall option in solaris. so
a reinstall of the OS will be required if you lose a machine.

It's still nice to know you can restore your data when you need it.

there's always some extra steps for backing up a live database though, but
this depends on the database itself.

The problem with stuff like dd is there's no granularity. If you want one
file, dd won't help you.


Ian Collins

unread,
Nov 6, 2012, 1:27:53 PM11/6/12
to
I forgot

6. Attach a spare disk as a mirror to the root pool, make it bootable,
split.

--
Ian Collins

Ian Collins

unread,
Nov 6, 2012, 1:39:30 PM11/6/12
to
On 11/07/12 07:20, Cydrome Leader wrote:
>
> there's not really any type of bare metal reinstall option in solaris. so
> a reinstall of the OS will be required if you lose a machine.

Not quite, you can boot from live media and recover your root pool and
any other data you have lost.

> It's still nice to know you can restore your data when you need it.
>
> there's always some extra steps for backing up a live database though, but
> this depends on the database itself.
>
> The problem with stuff like dd is there's no granularity. If you want one
> file, dd won't help you.

Quite. Snapshots are your friend.

--
Ian Collins

Drazen Kacar

unread,
Nov 6, 2012, 2:05:54 PM11/6/12
to
Cydrome Leader wrote:

> there's not really any type of bare metal reinstall option in solaris. so
> a reinstall of the OS will be required if you lose a machine.

I'm not sure I understand what you mean by "reinstall". My bare metal
recovery procedure is to boot the OS image from the network and that OS
image has backup client on it. And then you create file systems and
restore everything from the backup.

--
.-. .-. Yes, I am an agent of Satan, but my duties are largely
(_ \ / _) ceremonial.
|
| da...@fly.srk.fer.hr

Cydrome Leader

unread,
Nov 6, 2012, 5:59:15 PM11/6/12
to
Ian Collins <ian-...@hotmail.com> wrote:
> On 11/07/12 07:20, Cydrome Leader wrote:
>>
>> there's not really any type of bare metal reinstall option in solaris. so
>> a reinstall of the OS will be required if you lose a machine.
>
> Not quite, you can boot from live media and recover your root pool and
> any other data you have lost.

trying to rescue a machine is not full recovery.

if I set your server on fire and then throw it down the stairs, you're not
going to be recoving any root pools.

Cydrome Leader

unread,
Nov 6, 2012, 6:01:06 PM11/6/12
to
Drazen Kacar <da...@fly.srk.fer.hr> wrote:
> Cydrome Leader wrote:
>
>> there's not really any type of bare metal reinstall option in solaris. so
>> a reinstall of the OS will be required if you lose a machine.
>
> I'm not sure I understand what you mean by "reinstall". My bare metal
> recovery procedure is to boot the OS image from the network and that OS
> image has backup client on it. And then you create file systems and
> restore everything from the backup.

it's always a pita to reinstall solaris, no matter how you do it.


Ian Collins

unread,
Nov 6, 2012, 6:07:53 PM11/6/12
to
On 11/07/12 11:59, Cydrome Leader wrote:
> Ian Collins<ian-...@hotmail.com> wrote:
>> On 11/07/12 07:20, Cydrome Leader wrote:
>>>
>>> there's not really any type of bare metal reinstall option in solaris. so
>>> a reinstall of the OS will be required if you lose a machine.
>>
>> Not quite, you can boot from live media and recover your root pool and
>> any other data you have lost.
>
> trying to rescue a machine is not full recovery.
>
> if I set your server on fire and then throw it down the stairs, you're not
> going to be recoving any root pools.

Recover from backup, which may be a stream from another host.

--
Ian Collins

ChrisQ

unread,
Nov 6, 2012, 6:25:27 PM11/6/12
to
On 11/06/12 18:27, Ian Collins wrote:

>>
>> 4. Create a pool on spare disk(s) and zfs send your data there. If you
>> have space, rsync the data from the other OS systems there as well.
>>
>> 5. Commercial backup software!
>
> I forgot
>
> 6. Attach a spare disk as a mirror to the root pool, make it bootable,
> split.
>

I've recently tried snapshot zfs send to backup a root mirrored pool and
even wrote a short script to do it, but not sure if the bootblock is
saved as well. My guess is that it isn't ?. For data backup, I use rsync
to a second machine...

Regards,

Chris


Ian Collins

unread,
Nov 6, 2012, 7:44:15 PM11/6/12
to
On 11/07/12 12:25, ChrisQ wrote:
> On 11/06/12 18:27, Ian Collins wrote:
>
>>>
>>> 4. Create a pool on spare disk(s) and zfs send your data there. If you
>>> have space, rsync the data from the other OS systems there as well.
>>>
>>> 5. Commercial backup software!
>>
>> I forgot
>>
>> 6. Attach a spare disk as a mirror to the root pool, make it bootable,
>> split.
>>
>
> I've recently tried snapshot zfs send to backup a root mirrored pool and
> even wrote a short script to do it, but not sure if the bootblock is
> saved as well. My guess is that it isn't ?.

It isn't, but it's a one liner to restore it to the new drive.

I have just been testing recovery with Solaris 11.1 (the process is
documented here:
http://docs.oracle.com/cd/E26502_01/html/E29007/recover-4.html)

and I had a system back up and running on a new drive in about 10 minutes.

> For data backup, I use rsync to a second machine...

rsync is fine for non-Solaris systems, but snapshots are much more
flexible (you can recover individual files) and considerably faster to send.

--
Ian Collins

Ian Collins

unread,
Nov 6, 2012, 7:45:51 PM11/6/12
to
That depends how soft your arse is! See my reply to ChrisQ else-thread.

--
Ian Collins

BertieB...@gmail.com

unread,
Nov 7, 2012, 4:36:03 AM11/7/12
to
Looking at the options. This has got to be done as quick as possible.

Using dd would mean I'd have to physically swap in spare disk to every server (its 120 miles away). And I'd need a separate disk for each server.

flar is a possibility I guess - if we get failure then we've got the ability to do flar install I guess. Prob need to be single user to create though.

fssnap and ufsdump seems like a good option also. One thing is I cant take these servers down so its going to have to be in place backup.

Another idea - live upgrade to clone the disk?

Ian Collins

unread,
Nov 7, 2012, 4:53:32 AM11/7/12
to
What's wrong with sending snapshots?

--
Ian Collins

BertieB...@gmail.com

unread,
Nov 7, 2012, 6:18:35 AM11/7/12
to
U mean created with fssnap?

How easy is it to restore these then? Easier than ufsrestore?

John D Groenveld

unread,
Nov 7, 2012, 10:03:06 AM11/7/12
to
In article <186e46b6-bec2-42b1...@googlegroups.com>,
BertieB...@gmail.com <BertieB...@gmail.com> wrote:
>Got 5-6 Solaris 10 servers in remote location - power work being done.

Do you have bare metal spares to test your restores?

John
groe...@acm.org

Cydrome Leader

unread,
Nov 7, 2012, 11:57:57 AM11/7/12
to
so you stream data straight to a dead machine?


Ian Collins

unread,
Nov 7, 2012, 2:10:12 PM11/7/12
to
On 11/08/12 00:18, BertieB...@gmail.com wrote:
> On Wednesday, November 7, 2012 9:53:33 AM UTC, Ian Collins wrote:
>> On 11/07/12 22:36, BertieB...@gmail.com wrote:
>>
>>> Looking at the options. This has got to be done as quick as possible.
>>
>>> Using dd would mean I'd have to physically swap in spare disk to every server (its 120 miles away). And I'd need a separate disk for each server.
>>
>>> flar is a possibility I guess - if we get failure then we've got the ability to do flar install I guess. Prob need to be single user to create though.
>>
>>> fssnap and ufsdump seems like a good option also. One thing is I cant take these servers down so its going to have to be in place backup.
>>
>>> Another idea - live upgrade to clone the disk?
>>
>> What's wrong with sending snapshots?
>
> U mean created with fssnap?
>
> How easy is it to restore these then? Easier than ufsrestore?

**
Please remove all the crap that awful google interface adds to your replies!
**

I take it from your reply that the systems don't have ZFS root. I that
case, ufsdump is probably your best option.

--
Ian Collins

Ian Collins

unread,
Nov 7, 2012, 2:10:40 PM11/7/12
to
Very droll.

--
Ian Collins

"mr.fantać"

unread,
Nov 7, 2012, 5:58:05 PM11/7/12
to
You need to make disaster recovery flash image.
Procedure is easy to make but you need consider and set up your server
in that mode you can recover hole network with just one DVD.

1. If you use cluster than you need to backup root node
create flash image (flar create).
2. Set up NFS independet server which will contain SVR4 packages
3. Update your response file from Solaris DVD installation so when stage
2 is complete it runs automated installation and finds response file
with nfs servername from wherw it can install all required packages, of
course you need to create right profiles for each server or cluster node.
4. Good luck

Denis

BertieB...@gmail.com

unread,
Nov 8, 2012, 4:49:03 AM11/8/12
to
Nope. Unfortunately not.

John D Groenveld

unread,
Nov 8, 2012, 8:23:57 AM11/8/12
to
In article <36870b71-006a-4afb...@googlegroups.com>,
BertieB...@gmail.com <BertieB...@gmail.com> wrote:
>Nope. Unfortunately not.

What's your customer's plan should one or more systems fail?

John
groe...@acm.org
0 new messages