Topic: Backup to workstation WARNING!!

[Per Sørensen]

Hi

I am running SME 5.6 unsupportet + update4 on a PII 333 384Mb RAM with two 120G HD in a software RAID1

I encounteret two problems with the Backup to Workstation of which one is quite serious:

First, i wanted to do a backup of everything without ibays.

After watching the filefragment no. 6 at 700MB, i figured something was not quite right. It should not be more than one, so i found that when backing up /dev, it some how sees /dev/md1 as a very big file because off the RAID (i guess). Well, easy to fix: Exclude /dev from the BU, not really nessesary anywhay is it? BU went smooth after that and created a nice 214 MB file.

Next, i wanted to test the restore function.

I wanted to restore to a different dir just in case.. But i couldn't get it to work, it did opposite of what it should. Then i just tried a FULL restore, and it started OK, but then the server just went dead. COMPLETELY! I could not log in or anything. No response to ctrl-alt.del, so i had to switch the power of. Then it could not start because file; /lib/i686/libpthread-0.9.so was to short (0 bytes, i found later). I tried to do an upgrade with the CD - no luck. So i had to make a new install on another HD, mount one of the main ones and copy the file on to that then it bootet again.

So the question, is what can be restored and what can not. I am not in the mood for more testing, so does anyone have suggestions on what else to put in the Exclude list?

Per

[Ray Mitchell]

If you create the Disaster Recover backup job, and then look at the included folders, you will see what is absolutely essential to backup in order to be able to restore your server fully (to a functioning operating system).

It seems to me the best way to use backup2ws, is to use the 911 Disaster Recovery job as the basis of your backups (created say weekly), and then create other smaller selective backup jobs that backup other sets of folders at more regular intervals.

You can then do a major disaster recovery from the 911 backup and then add any more recent files from the selective backup jobs.

Of course you could use ONLY the 911 job and run it daily to backup everything that is changeable.

In my experience backing up your whole server is not the best idea. If you do do that, then I would only use this for archival purposes and restore only some of these files manually as needed, rather than a blanket "lets restore everything" approach.

Hope this helps
Regards
Ray

[Per Sørensen]

Ray Mitchell wrote:
>
> If you create the Disaster Recover backup job, and then look
> at the included folders, you will see what is absolutely
> essential to backup in order to be able to restore your
> server fully (to a functioning operating system).

Yes, it looks good for users and data and so.......
>
> It seems to me the best way to use backup2ws, is to use the
> 911 Disaster Recovery job as the basis of your backups
> (created say weekly), and then create other smaller selective
> backup jobs that backup other sets of folders at more regular
> intervals.

Agree


> In my experience backing up your whole server is not the best
> idea. If you do do that, then I would only use this for
> archival purposes and restore only some of these files
> manually as needed, rather than a blanket "lets restore
> everything" approach.¨

Actally i want to back up "everything" except the ibays, to be able to quickly get verything up running in case of a failure, hardware or human I also have a few things like Hylafax and Antivir installed, my idea was that if  "everything" except ibays was backed up, these things would be installed by restoring the BU on a fresh install. Will that be the case?

My point was also to point out that one should not backup things that crashes the server at restore. I don't know if this was a stand alone occurance or it will happen again. In my restore log file, the last file backed up was: /lib/i686/libpthread-0.9.so and then it crashed

Regards
Per

[Ray Mitchell]

> Actually i want to back up "everything" except the ibays,

Duplicate the folders backed up in the 911 job to another backup job.
Put the ibays folder in the excluded field
/home/e-smith/files/ibays

That should backup everything that is needed to restore a server to its prior functioning state except for the ibays data. (and also except for the add on packages)
 
I also have a few things like
> Hylafax and Antivir installed, my idea was that if
> "everything" except ibays was backed up, these things would
> be installed by restoring the BU on a fresh install.

I'm not absolutely sure how backup2ws & sme works in this regard, but in the event of a major failure requiring the OS to be reinstalled and backups restored, the recommended procedure is to  reinstal add in packages after the OS had been reinstalled.
I'm just in the process of testing this myself. So far it would seem to work OK from a 911 backup, but that does not include "everything".

Note that there are particular steps required to be performed when a 911 backup job is restored, there is another selection to make to run the system upgrade event (I think that's what it is).

> My point was also to point out that one should not backup
> things that crashes the server at restore.

I'm wondering whether the "everything" restore actually broke the server, ie stopped system files from functioning as they were replaced ?? I don't know enough about the processes involved to analyse it further.

Sounds like you perhaps might want to temporarily remove your ibay data and then create a Ghost image of your drive, that way you could just reinstate a "system without ibay data" and then copy the ibay data back on.

Hope this helps
Regards
Ray Mitchell

[Per Sørensen]

Ray Mitchell wrote:
 
> I'm not absolutely sure how backup2ws & sme works in this
> regard, but in the event of a major failure requiring the OS
> to be reinstalled and backups restored, the recommended
> procedure is to  reinstal add in packages after the OS had
> been reinstalled.
> I'm just in the process of testing this myself. So far it
> would seem to work OK from a 911 backup, but that does not
> include "everything".

I will look forward to hear about your test. If i set up a server in the near future, i will try myself.
 
> > My point was also to point out that one should not backup
> > things that crashes the server at restore.
>
> I'm wondering whether the "everything" restore actually broke
> the server, ie stopped system files from functioning as they
> were replaced ?? I don't know enough about the processes
> involved to analyse it further.

Me too.....
 
> Sounds like you perhaps might want to temporarily remove your
> ibay data and then create a Ghost image of your drive, that
> way you could just reinstate a "system without ibay data" and
> then copy the ibay data back on.

Now that's a good idea. I just have to do some juggling with the hard drives since i'm running software RAID1.

Thanks for the help.
Per

> Hope this helps
> Regards
> Ray Mitchell

It sure does!

[Guck Puppy]

Ray Mitchell wrote:

> > I also have a few things like
> > Hylafax and Antivir installed, my idea was that if
> > "everything" except ibays was backed up, these things would
> > be installed by restoring the BU on a fresh install.

> I'm not absolutely sure how backup2ws & sme works in this
> regard, but in the event of a major failure requiring the OS
> to be reinstalled and backups restored, the recommended
> procedure is to  reinstal add in packages after the OS had
> been reinstalled.

As a rule, whenever I have installed any 3rd party rpm's, I work out of a created "rpm" dir in root's homedir. In that dir is a subdir called "installed". I anticipate being able to do:

rpm -Uvh rpm/installed/*

when the time comes to reinstall fresh or whatever. Granted, there are one or two special case rpm files that need a little more prep and special installation ordering, but I note that whenever it arises.

G

[Upaboveit]

Per Sørensen wrote:
>

> After watching the filefragment no. 6 at 700MB, i figured
> something was not quite right. It should not be more than
> one, so i found that when backing up /dev, it some how sees
> /dev/md1 as a very big file because off the RAID (i guess).
>

Not because of the RAID. Same thing happened to me. I have a 36GB SCSI drive in the box with about 3GB of data. After watching 650MB .rar file number 15 being generated something was obviously wrong. Sure enough most of the .rar files were of sda. Seems its trying to backup everything, including free space.