cancel
Showing results for 
Search instead for 
Did you mean: 

Difference Between Service Group Disable & Offline

Rajan_Iyer
Level 4
Certified

Hi,

 

I would like to the difference between Service Group Disable & Offline.

In the same ref. I have below queries.

1. What is the difference between them.

2. During performing both i.e. what need to be first i.e. first Disable & Offline.

3. After performing above options if I will reboot the server whether the service group will remain Disabled & Offline ?

4. In case of making it online I think first I need to enable the Service Group & online it ----is it correct ?

 

Please suggest.

 

Thanks

 

6 REPLIES 6

Marianne
Moderator
Moderator
Partner    VIP    Accredited Certified

Offlining a Service group will offline all resources, bu NBU will still monitor them If, for example someone brings a resource online from OS (e.g import diskgroup and mount filesystem) VCS will report those resources as online at the next monitor cycle.

Disabling the SG stops monitoring the resources in the SG and the SG cannot be brought online. Only when SG is enabled can the resources in the SG be brought online.

See the following : https://sort.symantec.com/public/documents/sf/5.1/solaris/html/vcs_admin/ch_admin_vcs_from_cli59.html

If you want to prevent SG from coming online after reboot, best to do persistent freeze of the SG..

mikebounds
Level 6
Partner Accredited

Bascially disabling a service group means VCS ignores its there so you might use this if you created SG before installing application.  Freezing SG, stops VCS taking any action - it does nothing to the resources so leaves them in the same state (ONLINE or OFFLINE) but it still monitors the resources. 

Offlining a service group offlines the resources (i.e take the App down).

Mike

Anoop_Kumar1
Level 5

 

SG Disable : SG will not be able to online and failover
 
SG Autodisabled : SG state is unknown to Cluster due to resources not probed or VCS not running on visible server in cluster.
 
Offline : The resources are in Offline State and are monitored by VCS for any state change of resources in cluster.
 
 
Generally, nobody disabled the SG. To avoid failover actions by VCS in case of fault on resources in SG, one use to Freeze the SG. A persistent Freeze allows freezing among reboots too.

kjbss
Level 5
Partner Accredited

I wasn't in agreement with several of the comments here, so I just tested this on a Solaris system running "5.0_REV-5.0MP3" -- which I admit is old, so it would be interesting to hear how 5.1 and 6.0 behave...

 

As Anoop says, "SG Disable : SG will not be able to online and failover", this is technically true.  you cannot "hagrp -online" nor "-switch" nor will the SG failover in the event of critical resource faulting whilst the SG was in the disabled "condition". 

However, you CAN bring the SG "online" by manually onlining  (outside VCS) all the resources in the SG.  I accept that this could be considered "cheating".  None the less, a disabled SG can be made to go online.

 

Now on to my main point:

This proves that a "Disabled" SG *does* still monitor all it's enabled resources, every OfflineMonitorInterval, which may be surprising to many.  Moreover, once online, and you manually offline the resource (outside VCS), the Monitor entry point discovers it's faulted, and the clean entry point is called, and the "Disabled" SG goes to the "OFFLINE|FAULTED" state.  (One presumes that this would obey the current value of ManageFaults, that is, if ManageFaults=NONE, then clean would not have been called). 

 

I *think* there may be a case for suggesting that doing "hagrp -modifiy Enable 0 SG", aka: disabling a SG, should also disable all enabled resources in the SG, but that might be a step too far?  Personally, I think it should. 

So, the way it works now (err, at least on 5.0_REV-5.0MP3) seems to be vary similar to a froozen SG.  Can anyone say what the difference is between a froozne SG and a disabled one?

 

Sorry if this is *old* news and has been changed on more recent levels.. (but please do let me know!)

g_lee
Level 6

----------[ EDIT ]----------

Comments below are NOT correct - incorrect output as group was reenabled while testing and I didn't realise until after kjbss replied. Have left post as is, otherwise kjbss' reply makes no sense. Apologies for any confusion cause by this error.

----------[ end EDIT ]----------

kjbss,

You are correct that individual resources can be brought online in a disabled group; the point of the disable is that group operations are disabled, eg: you cannot run:

# hagrp -online <group> <-any|-sys {sys}>

or

# hagrp -switch <group> ....

As you said, you can disable each individual resource in the group as well, but this is separate to disabling the group.

regarding differences between disabling and freezing a group - if you freeze a group (rather than disabling), this also prevents operations on the resources in the group

ie:: if GroupA is disabled, you can still online individual resources in the group within VCS (as you discovered), whereas if GroupA is frozen, if you try to run hares -online <res>, it will issue the following warning:

VCS WARNING V-16-1-10276 Group <group> for resource <resource> is frozen

kjbss
Level 5
Partner Accredited

g_lee says: "ie:: if GroupA is disabled, you can still online individual resources in the group within VCS (as you discovered)"

 

No, no; that is hopefully not what I said, because you cannot use the VCS interfaces (CLI, GUI, etc.) to online a resource that is in a disabled SG:

$ hagrp -display kjb0 | egrep -i 'Enabled'

kjb0         Enabled               sys1     0
kjb0         Enabled               sys2     0

$ hares -online kjb0_1_foo -sys sys2
VCS WARNING V-16-1-10280 Group kjb0 for resource kjb0_1_foo is not enabled on system sys2

 

...and therefore, the 5.0_REV-5.0MP3 implementation of VCS seems to treat a disabled SG and a "froozen persistent" SG about the same.  I cannot really see the difference between the two (yet).  Well, except by default the Frozen SG attributes are "global" and the Enabled SG attribute is local. 

As I said, I kinda think that if you disable a SG (hagrp -disable <group>), it ought to automatically disable all of the contained resources, as if you did "hagrp -disableresources <group>" directly after disabling the SG.  I guess if YOUR intent is to not only disable the SG, but to also disable the resources in the SG (to stop them from monitoring), then YOU need to remember to disable the resources too. 

I would still like to hear if this behaves the same for 5.1 and 6.0... anybody got time and access to systems to do the test and let us know?