Backup selection for Linux clients

[Amaan]

Please help, i am little bit confused about backup selection for Linux clients. My current policy was ALL_LOCAL_DRIVES and follow NFS and cross mount point attributes checked.

I read in the admin guide that cross mount point should be uncheck while backup selection is ALL_LOCAL_DRIVES.

Quetion is Why? what is the impact?

Next question is what should i do to get the mount points to be backed up? in the job details NB missing the as different mount point if i will do just ALL_LOCAL_DRIVES without cross mount point.

Please help on this.

Thanks!

[Yasuhisa_Ishika]

ALL_LOCAL_DRIVES takes all local drives itself, and you don't have to check Cross Mount Point.
If checked, it is possible that mounted volumes are backed up twice. Once, each volumes are backed up by ALL_LOCAL_DRIVES. And each volumes will be backed up twice because each volumes mounted under root filesystem may be followed by backup of root file system. To avoid this, you should uncheck Cross Mount Point.

> Next question is what should i do to get the mount points to be backed up?
> in the job details NB missing the as different mount point if i will do just
>  ALL_LOCAL_DRIVES without cross mount point.

Where did you determine NetBackup did not take backup of non-root FS?
Have you already check backup contents by BAR or bplist command?
 

[Marianne · Accepted solution]

ALL_LOCAL_DRIVES will backup all local mountpoints of supported filesystem types, such as UFS, VxFS, ext3.

Filesystems that are automatically excluded when backing up / or ALL_LOCAL_DRIVES are listed in this TN:

http://www.symantec.com/docs/TECH52350

I personally prefer ALL_LOCAL_DRIVES (without cross mount points and follow NFS) with Allow Mutiple Data Streams selected.  If databases on clients, data files/directories need to be added to exclude_list.

[Amaan]

I just tested, my backup selection was ALL_LOCAL_DRIVES and cross mount point and follow nfs disabled and multistreaming enabled as suggested by Marianne. In the screenshot (attached) i am not able to get the files for /home or /globe and etc...

and in the job details i am seeing following:

2/23/2012 9:12:59 AM - Info bpbrm(pid=6156) from client 3: TRV - /opt/global is in a different file system from /opt. Skipping.
2/23/2012 9:12:59 AM - Info bpbrm(pid=6156) from client 3: TRV - /opt/home is in a different file system from /opt. Skipping.
2/23/2012 9:13:39 AM - Info bpbrm(pid=6156) from client 3: TRV - /opt/hpservices/adm/.serverSocket is a socket special file. Skipping.
2/23/2012 9:13:51 AM - Info bpbrm(pid=6156) from client 3: TRV - /opt/eco is in a different file system from /opt. Skipping.

2/23/2012 9:10:36 AM - Info bpbrm(pid=7300) from client 3: TRV - /usr/local/unicenter is in a different file system from /usr. Skipping.
2/23/2012 9:11:33 AM - Info bpbrm(pid=7300) from client 3: TRV - /usr/openv is in a different file system from /usr. Skipping.

are those really skipping? if i want them to restore where i will find them? Are they being backed up in different directory?

[CRZ]

They're skipping because, as the log says, they're on different file systems.  Presumably they would be picked up by a different stream but you wouldn't see that in the Job Details.

But let's learn some more about your environment to properly address your issues:

Nowhere have you said what version of NetBackup you're running.  What's the NetBackup version of the servers AND clients involved?

Which Linux exactly are these clients (name/version)?  Can you provide `uname -a` output or redhat-release contents (if applicable)?

How about the output of a `bpmount -i` command so we can see what kind of file systems we're discussing for particular mount points?

[Yasuhisa_Ishika]

This screenshot shows empty /global directory.
Please check /opt/global.
Also check your missing mount point is NFS mount or local mount.

[Marianne]

Please post output of 'df -h' on the client.

Are you 100% sure there is a /home and /global filesystem?

According to the job details the client has /opt/global and /opt/home filesystems.

So, while a job is running for /opt, those messages are normal (/opt/home is in a different file system from /opt. Skipping.) You should be able to see separate jobs for /opt/global and /opt/home.

Same for the separate mount points under /usr - they will be backed up on their own.

[Amaan]

Sorry for late answer. I dont have access to the client so i cannot give an answer to most of your questions.

I tried as Marianne said, but i got this for one of the streams which had backup selection "/"

2/24/2012 11:52:55 AM - Info bpbrm(pid=5112) from client om: TRV - /stand is in a different file system from /. Skipping.
2/24/2012 11:52:56 AM - Info bpbrm(pid=5112) from client om: TRV - /tmp is in a different file system from /. Skipping.
2/24/2012 11:52:56 AM - Info bpbrm(pid=5112) from client om: TRV - /backupex is in a different file system from /. Skipping.
2/24/2012 11:52:56 AM - Info bpbrm(pid=5112) from client om: TRV - /opt is in a different file system from /. Skipping.
2/24/2012 11:52:56 AM - Info bpbrm(pid=5112) from client om: TRV - /usr is in a different file system from /. Skipping.
2/24/2012 11:52:57 AM - Info bpbrm(pid=5112) from client om: TRV - /var is in a different file system from /. Skipping.
2/24/2012 11:52:57 AM - Info bpbrm(pid=5112) from client om: TRV - /syslogs is in a different file system from /. Skipping.
2/24/2012 11:53:12 AM - Info bpbrm(pid=5112) from client om: TRV - /dev/log.un is a socket special file. Skipping.
2/24/2012 11:53:12 AM - Info bpbrm(pid=5112) from client om: TRV - /home is in a different file system from /. Skipping.
2/24/2012 11:53:12 AM - Info bpbrm(pid=5112) from client om: TRV - /net is filesytem type autofs. Skipping.
 

I found all of the "skipped" ones another seperate job, but couldnt find for /net

Then i checked follow nfs in the policy and start the job again, but this time I got this for /net

/net is in a different file system from /. Skipping.

Why i was not able to get this backed up?

[J_H_Is_gone]

after doing a backup of all local drives - have you then looked in the BAR and looked at the backup images to see what was backed up.

unix is not like windows.

in unix / (known as root) is a filesystem (think if it like a drive called / )

then you have other 'filesystems'  like home and opt

in unix they are "mounted" on top of  root  - so the other filesystems look like /home  and /opt

so if you backed up everything in / you would get root and home and opt

but as you are backing up all local drives  /home and /opt are really different filesystems (like different drives)  and will be backed up as their own.

And example would be if you have multiple streams enabled you would get

one job for /

one job for /home

one job for /opt

when you go look in the BAR at the backup images take you will see

/

  home

   opt

----

now what you are seeing in that one job when it says

 /opt is in a different file system from /. Skipping.
 

this is just saying that at this time it is backing up / (root) and sees that opt is mounted on / - but it knows it is a different filesystem so it will get it when it backups the file system opt - it is not going to back it up while backing up / because then it would back it up twice.

So go check the BAR - look at your backup - check to see if it really skipped any of those filesystems.

[CRZ]

It sounds like you've figured it all out except for /net.  This one's simple as it's stated right in the Job Details:

2/24/2012 11:53:12 AM - Info bpbrm(pid=5112) from client om: TRV - /net is filesytem type autofs. Skipping.

I would guess that if you COULD give us `df -h` output, you'd find /net mounted as something like "-hosts" as opposed to a directory that could be backed up.  (I don't think NetBackup backs up autofs at all, does it?  I could be totally wrong here)

[Marianne]

If I am not mistaken, /net is also an NFS mount (which is getting skipped by default).

Extract from  http://www.centos.org/docs/5/html/Deployment_Guide-en-US/s1-nfs-client-config-autofs.html :
The automount utility can mount and unmount NFS file systems automatically (on demand mounting)....

Also http://www.vanemery.com/Linux/NFS-Van.html : 
With the automount utility (a.k.a. autofs), NFS directories can be mounted and unmounted automatically as needed by regular users.

[Amaan]

According to all above i want to make a conclusion. Please correct me if I misunderstood something.

• Best selection to get all files backed up is ALL_LOCAL_DRIVES and cross mount point is unchecked and follow NFS is unchecked;
• Enabling multistreaming is prefered but not neccessary;
• Even if they are saying that it is skipped it will be backed up (it is visable when you enable multistreaming);
• /net is skipped because it is the file (location, path) which will be skipped by default.

I didnt se the difference in my backups while enabling Follow NFS and disabling it. Does it hurt if it is enabled (checked) all the time?

[js88699]

You have had lots of good advive by very good people.  Lets go over your last post.

Best selection to get all files backed up is ALL_LOCAL_DRIVES and cross mount point is unchecked and follow NFS is unchecked;   This is considered by most to be the best practice.  This will backup ALL local mount points on UNIX.  If you use multi-stream, you will see a job for each one.  If you don't use multi-stream, you won't.  The mount points will not show up in the Job Overview, just ALL_LOCAL_DRIVES.
 

Enabling multistreaming is prefered but not neccessary;  Correct, but I have found it easier to troubleshoot a failed job if you know what mount point or windows drive it was trying to do.  Personal preference.

Even if they are saying that it is skipped it will be backed up (it is visable when you enable multistreaming); That is also correct, you will see lots of skipped messages for the / filesystem.  If you have other mount points stacked on top of each other, you will see it there.  As you correctly summised, you will see each mount point in it's own job stream if you have multi-stream enabled.

/net is skipped because it is the file (location, path) which will be skipped by default.  It sounded like /net was a NFS filesystem, so unless this is the only place it's backed up from.  I would suggest you leave NFS unchecked.  That filesystem should be backed up from the system that is sharing it, not the ones that use it.

It looks like you are starting to get the hang of things.  If you use bplist or go into the Backup, Archive, and Restore part of the gui you can look and see what was backed up for that client.

[Amarnath_Sathis]

When you select Follow NFS,

The Backup data will be more than the actual data reside on the server.