Monitoring file systems backup Post: 302901048

Sponsored Content

Special Forums UNIX and Linux Applications Infrastructure Monitoring Monitoring file systems backup Post 302901048 by hicksd8 on Sunday 11th of May 2014 07:45:52 AM

05-11-2014

Moderator

You don't state what hardware platform you have, what the cluster software suite is, or what the backup software is.

Your post indicates that you have a good understanding of how a (generically speaking) cluster works and that any one filesystem can only be under the control of one node at a time. Having multiple nodes thinking they could write to the volume would be anarchy and a clear recipe for data corruption. It is definitely the job of the cluster software suite to ensure that that never happens. Having said that, different cluster suites can have starkly different functionality.

Similarly, backup software suites also vary in the manner of operation.

So discussing cluster backup in generic terms I would say that there are two options for implementing backups. Firstly, when node-A fails and node-B takes over (by checking orphaned filesystems and then mounting them, taking over and broadcasting the cluster name and ip address (node-C and ipaddr-C) some cluster software will also failover scheduled jobs (eg, backup). Of course, the backup device(s) need to be still available (or node-B needs to have its own tape drive, for example) for this to work. Alternatively, like all the user community who only know about node-C and ipaddr-C, the backup is run from a machine outside the cluster which "calls in" on node-C, accesses or NFS mounts the filesystem, and backs it up. Usually, this is the preferred method.

Now in this scenario the backup software has no knowledge that it is backing up a cluster volume and it should work exactly the same way as it would with a local volume, ie, if it loses communication with the volume, it will report a backup failure. Some backup software suites (eg, NetBackup) are of client/server architecture which are very intelligent and will report failures in exactly the same way they usually do.

So in summary, the fact that it is a cluster should be largely irrelevant to reporting errors in backup schedules. How the success of a backup is verified is the same as the non-cluster scenario.

Hope that helps. Feel free to continue your questions but please give us all a clue of the platform and software(s) involved.

Last edited by hicksd8; 05-11-2014 at 12:46 PM..

hicksd8

View Public Profile for hicksd8

Find all posts by hicksd8

4 More Discussions You Might Find Interesting

1. Filesystems, Disks and Memory

Backup Linux file systems

Hi, I need to backup files on our Solaris machines onto a windows machine. I have Samba installed. Is it possible to backup these files on Solaris/Linux machines onto a Windows machine. Thanks Aravind

2. UNIX for Dummies Questions & Answers

Backup Software for UNIX systems

Hi, which software is recommended for backup of UNIX systems ( e.g. SUN, Solaris ). Backup software and database e.g. Oracle. One possibility is Networker, but license is expensive and also service contract. Best regards Dieter

3. UNIX for Dummies Questions & Answers

Monitoring Unix systems

I am looking for a commercial tool that will give me -UNIX Monitoring performance solution+ reports on CCV format. (as perfmon on windows machines). The tool must have following counters per PROCESS: Page Faults/sec Virtual Bytes % Processor Time Handles count Threads count...

4. AIX

How to exclude directory from (File Systems) backup?

Hello AIX experts, I have a file system called /bossapp Its size = 77.5 GB I want to take a File Systems backup for this one using smitty, it is very easy, but the problem is I want to exclude one directory called (ORIGIN). How? The steps are very easy to take a File Systems backup,...

LEARN ABOUT HPUX

cmhaltnode

cmhaltnode(1m)															    cmhaltnode(1m)

NAME

       cmhaltnode - halt a node in a high availability cluster

SYNOPSIS

       cmhaltnode [-f] [-v] [-t] [node_name...]

DESCRIPTION

       cmhaltnode causes a node to halt its cluster daemon and remove itself from the existing cluster.

       To halt cluster on the node, a user must either be superuser(UID=0), or have an access policy of FULL_ADMIN allowed in the cluster configu-
       ration file. See access policy in cmquerycl.

       When cmhaltnode is run on a node, the cluster daemon is halted and, optionally, all packages that were running on that node  are  moved	to
       other nodes if possible.

       If node_name is not specified, the cluster daemon running on the local node will be halted and removed from the existing cluster.

       If  you issue this command while a cluster is still in the process of forming, the command will fail with the message "Unable to connect to
       daemon."  If this happens, wait for the cluster to form successfully, then issue the command again.

   Options
       cmhaltnode supports the following options:

	      -f	Force the node to halt even if there are packages or group members running on it.  The group members on the node  will	be
			terminated.   The halt scripts for all packages running on the node will be run; based on priority or dependency relation-
			ships, this may affect packages on other nodes. In other words, packages on other nodes may either start or halt based	on
			this  package halting.	If the package configuration and current cluster membership permit, and if the package halt script
			succeeds, the packages will be started on other nodes.	Without this option, if packages are running on  the  given  node,
			the command will fail.	If a package fails to halt, the node halt will also fail.

	      -v	Verbose output will be displayed.

	      -t	Test  only.  Provide  an assessment of the package placement without affecting the current state of the nodes or packages.
			This option validates the node's eligibility with respect to the package dependencies as well as the external dependencies
			such as EMS resources, package subnets, and storage before predicting any package placement decisions. If there is a pack-
			age in maintenance mode running on the nodes being halted, the package will always be halted and not failover  to  another
			node; the report will not display an assessment for that package.

	      node_name...
			The name of the node(s) to halt.

RETURN VALUE

       cmhaltnode returns the following value:

	       0   Successful completion.
	       1   Command failed.

EXAMPLES

       Halt the cluster daemon on two other nodes:

	      cmhaltnode node2 node3

AUTHOR

       cmhaltnode was developed by HP.

SEE ALSO

       cmquerycl(1m), cmhaltcl(1m), cmruncl(1m), cmrunnode(1m), cmviewcl(1m), cmeval(1m).

						      Requires Optional Serviceguard Software					    cmhaltnode(1m)