Different retention periods

joluinfante

ADSM.ORG Member
Joined
Sep 28, 2015
Messages
22
Reaction score
0
Points
0
PREDATAR Control23

Hi!
I'm using tsm4ve.
Is it possible to have...?

- Two different VM backup schedules defined in the GUI (for two arbitrary groups of VMs, on the same datacenter)
- Different retention periods (in days) for each group / schedule.
- Share the storage pool for those two backups.
- Use sched mechanism to make those backups (instead of launching the "dsmc backup vm" command from external cron).

TIA
jorge infante
rosario - santa fe - argentina
 
PREDATAR Control23

Yes, but you don' t need 2 schedules to use a different management class (retention). You would only need 2 schedules if you want to run them at separate times.

So, if you need to run two groups at different times, then the method to this is different than if you want to run one schedule that does what you want.

With 2 schedules, you would need 2 option files, one for each scheduler. You will need to add VMMC in the option file to specify the management class.

If you stick to 1 scheduler, you will need 7.1.x and use something like this in the option file:
Code:
include.vm vm1 mc1
include.vm vm2 mc1
include.vm vm3 mc1
include.vm vm4 mc2
include.vm vm5 mc2
include.vm vm6 mc2
 
PREDATAR Control23

thanks for the reply. I still have problems, surely, for lack of use of the tool. for now, I'm just doing backups of virtual machines to disk. I have in mind to handle archives. Initially, I backup all virtual disk with the following definitions:

define domain rm
define policy rm rmvms
define mgmt rm rmvms rmvms_daily
define devclass rmvms_daily devtype=file format=drive maxcapacity=100G
define stgpool rmvms_daily rmvms_daily pooltype=primary maxscratch=200 deduplicate=yes collocate=no
define vol rmvms_daily /tsm-data/data/rmvms_daily_v01 F=102400 N=10
define copygroup rm rmvms rmvms_daily type=backup dest=rmvms_daily verexists=nolimit verdeleted=nolimit retextra=90 retonly=90
assign defmgmtclass rm rmvms rmvms_daily
validate policy rm rmvms
activate policy rm rmvms

this definition went perfectly, but the problem is that I finished the disk space.
then, we decided to separate backups of vm's with no user data (eg, terminal servers or application servers) of vm's with user data (eg fileservers).
and give a "small" retention to vm's without user data and a "long" retention to vm's with user data.
for this, I did the following configurations:

define policy rm rmvmcfg_pol # for vm's without user data
define policy rm rmvmdat_pol # for vm's with user data
define mgmt rm rmvmcfg_pol rmvmcfg_mc
define mgmt rm rmvmdat_pol rmvmdat_mc
define copygroup rm rmvmcfg_pol rmvmcfg_mc type=backup dest=rmvms_daily verexists=nolimit verdeleted=nolimit retextra=14 retonly=14
define copygroup rm rmvmdat_pol rmvmdat_mc type=backup dest=rmvms_daily verexists=nolimit verdeleted=nolimit retextra=14 retonly=14
assign defmgmtclass rm rmvmcfg_pol rmvmcfg_mc
assign defmgmtclass rm rmvmdat_pol rmvmdat_mc

the destination of all copygroups will be my actual storage
i'd like, when i executed the next backup, tsm will apply the news retentions

then:
validate policy rm rmvmcfg_pol

I'm getting this message:
ANR1150W Management class RMVMS_DAILY is not defined in policy set RMVMCFG_POL but is defined
in the active policy set for domain rm: files bound to this management class will be redound to the default management class if/when this set is activated

What is the problem?
I don't like a UNIQUE management class, because the mc well define the copygroup, and the copygroup define the retention.

You can aclarme or correct me this idea?
 
PREDATAR Control23

excuse me. in the second copygroup the correct definition is:

define copygroup rm rmvmdat_pol rmvmdat_mc type=backup dest=rmvms_daily verexists=nolimit verdeleted=nolimit retextra=90 retonly=90
 
PREDATAR Control23

I did write this dsm.sys:

When I did run the rmvms_daily sched, I did receive this messages:
ANS9388E The management class RMVMCFG_MC specified for the VMCTLMC option is invalid, or does not have a backup copy group.

servername dm1_mitre
nodename dm1_mitre
commethod tcpip
...
vmctlmc rmvmcfg_mc
include.vm vm1 rmvmdat_mc

I'd like to use rmvmcfg_mc for a lot of vm's (for default).
And, use the rmvmdat_mc for a few machines with user data.
What is the error?

TIA
jorge infante
 
PREDATAR Control23

Is the node in the RM domain? (q node)
Does the RM domain have a management class named RMVMCFG_MC? (q mgmtclass rm active)
 
PREDATAR Control23

Yes, the node is on the rm domain:
q node:

Node Name Platform Policy Domain Days Since Days Since Locked?
Name Last Acce- Password
ss Set
------------------------- -------- -------------- ---------- ---------- -------
DM1_MITRE TDP VMw- RM <1 49 No
are
RM TDP VMw- RM <1 49 No
are
RMCLI TDP VMw- RM <1 49 No
are
RM_MP_LNX Linux RM 2 49 No
x86-64
RM_MP_WIN TDP VMw- RM 2 49 No
are
VC-MITRE (?) RM 49 49 No


q mgmtclass rm active:

Policy Policy Mgmt Default Description
Domain Set Name Class Mgmt
Name Name Class ?
--------- --------- --------- --------- ------------------------
RM ACTIVE RMVMCFG_- Yes
MC
RM ACTIVE RMVMDAT_- No
MC

I did resolve my above problems with:
define policy rm rmvms_pol
define mgmt rm rmvms_pol rmvmcfg_mc
define mgmt rm rmvms_pol rmvmdat_mc
define copygroup rm rmvms_pol rmvmcfg_mc type=backup dest=rmvms_daily verexists=nolimit verdeleted=nolimit retextra=14 retonly=14
define copygroup rm rmvms_pol rmvmdat_mc type=backup dest=rmvms_daily verexists=nolimit verdeleted=nolimit retextra=60 retonly=60
assign defmgmtclass rm rmvms_pol rmvmcfg_mc
validate policy rm rmvms_pol
activate policy rm rmvms_pol

One policy, two mc for retention periods.
And the options in the dsm.sys:
VMMC RMVMCFG_MC * default management class, lower retention
include.vm miranda RMVMDAT_MC * specific mc, higher retention
include.vm svn-server RMVMDAT_MC * specific mc, higher retention

But, in the first backup with this configuration (previous restart of dsmcad daemon), I did lost the version backups for "miranda" and "svn-server" greather than 14 days. Any explanation for this?
 
PREDATAR Control23

No explanation, only guesses. While trying to setup up the policy, it's possible that before you created the new management classes, the default had a shorter retention and you backed up to that one. Again, only a guess.
 
PREDATAR Control23

When you say "While trying to setup up the policy", you say:
* I did set all configuration of new management class.
* I did modify the dsm.opt file, with new mc assigments (I did try to reconfigure, and learn it in the process)
* I did have errors on this, and, with this errors, all data get incorrect mc assigments, dropping old versions.

My question is:
Where is the information about retention of data?
On each data? What is the method to check it? I'd like see this information about is object. There is exist?
I can't see, in the logs of backup, information about "I'm processing this job of backup to backup this object using this retention".
Another question is: What is the automatic process to apply the "data for delete because it is out of retention" mark, and do it? When this process run? Where are the logs of this process?

Thanks a lot for all the help you're giving me!
 
PREDATAR Control23

Can't really answer question by question, because it won't make sense.

The retention is in the copygroup for the management class the object is bound to in the domain the node belongs to. The retention is not with the actual object, the object is just bound to a management class.

If you change a node to a new domain and the management class from the old domain doesn't exist in the new, then the retention of the default management class takes effect.

You can use the query vm command to see which management class it is bound to right now. But that won't tell you if that management class existed all along.

Every time the expiration runs on the server, it looks at the objects and compare it with the current management classes and current retention. The expiration is logged in the activity log, but it will just tell you for which node and filespace it's running and how many objects it expired. No information about individual objects otherwise you'd need a few terabytes of storage just for the activity log.

So, with the problems you have been having, if the expiration ran once or multiple times before the configuration was correct, it's possible that it expired data earlier than you would have liked because if the management class did not exist, it would have used the default which could be shorter.

It will be hard to find the exact reason now, and it would not be of much value because once expired, it's gone.
 
Top