setting growth limit vs. growth warning

[mugurs]

Hi All,
I understood that for a CPG that is part of an AO config you have to set "growth warning" and not "growth limit". But what if the free space on CPGs with SSD (let's say) starts to shrink because of a tunesys command that is running and cannot be stopped. I saw that the allocated capacity grew more than the "growth warning" and I thought maybe if I had set up the "growth limit" that would have appeared.

Thanks,
mugurs

[Davidkn]

Page 12 of the best practice guide suggests that you should not set any growth limits, only growth warnings.

http://h20195.www2.hp.com/V2/GetPDF.asp ... 524ENW.pdf

[Davidkn]

In the AO section of the same document it says the following

Best practice: It is preferable to not set any capacity limit on the Adaptive Optimization configuration level, or on the CPG (no allocation warning or limit).
This will allow the Adaptive Optimization software to make excellent use of the different tiers available in the system. In HP 3PAR OS version 3.1.2 and later, use the MC to set the CPG capacities.
If a capacity limit is required for a given tier, set a capacity-warning threshold (not limit) on the CPG itself through the Management Console or the CLI. Adaptive Optimization will not attempt to move more data to a CPG than the capacity warning set on the CPG.

[mugurs]

I know what the "best practice" teaches us. Unfortunately, there are some specific critical situations when "best practice" can't help you. For a CPG where I've set only "growth warning" I've got allocated capacity bigger than the respective capacity specified by "growth warning". And that allocated capacity grew bigger and bigger due to an unstoppable tunesys and I thought in this case would have helped to set "growth limit" and not "growth limit". it appeared to me that "growth warning" had not any effect on limiting the allocated capacity.

thanks,
mugurs

[hdtvguy]

I do not believe growth warning is used by anything other than AO and mechanisms to alert you a CPG has exceeded the warning limit. If tunesys ran away there is nothing you can do other than to re-tune or wait and see if AO eventually clear out some space. AO will NOT move data out of a CPG above the warning just to get it below the warning, but the natural order of blocks migrating will eventually accomplish the same thing. We had the same issue, a CPG got too large and was 48TB and we set the warning to 40TB. Over a week or 2 as AO ran every day data would not be moved into that CPG, but data in that CPG that no longer behaved according to AO for that CPG was moved out and the CPG eventually got under the growth warning.

[Davidkn]

I know what the "best practice" teaches us. Unfortunately, there are some specific critical situations when "best practice" can't help you. For a CPG where I've set only "growth warning" I've got allocated capacity bigger than the respective capacity specified by "growth warning". And that allocated capacity grew bigger and bigger due to an unstoppable tunesys and I thought in this case would have helped to set "growth limit" and not "growth limit". it appeared to me that "growth warning" had not any effect on limiting the allocated capacity.

thanks,
mugurs

I'm not quite sure I understand what you're saying, but setting a limit will surely at some point stop io from writing to the cpg, hence why hp do not recommend using it?

And best practice is normally there as a recommendation to stop you from getting into a 'critical' situation.

I agree it's not always the final word, but it's normally a good indication of what you should choose.