Ask The Experts

LS_Mirrors & SVM_DR Best Practices

brasseurm

Hello all,

 

I have some questions regarding the usage of LS_Mirrors in a SVM_DR configuration.  I've read it is a Best Practice to use LS_Mirrors to protect root volumes in a NAS environment.  My system is a 4 nodes cluster, using only FlexVol and with CIFS SVM_DR.

  • Is it "mandatory" or really "recommanded" to put one a on each 4 nodes of the cluster, since the aggregates where the volumes reside are owned by a single HA Pair ?
  • What's your recommandation regarding snapshots of these LS_Mirror volumes ?
  • What about performance, is it really worth to use it ?
  • Do you have any example of circumstance when you can lose a root volume and, thus promote a LS_Mirror ?

 

Thank you for your help, do not hesitate to provide me any other guideline about this...

 

 

2 REPLIES 2

Re: LS_Mirrors & SVM_DR Best Practices

GidonMarcus
  • Is it "mandatory" or really "recommanded" to put one a on each 4 nodes of the cluster, since the aggregates where the volumes reside are owned by a single HA Pair ?
    • highly recommend
  • What's your recommendation regarding snapshots of these LS_Mirror volumes ?
    • no need to. you can't restore them or the root volume itself. you can only create new ones.
  • What about performance, is it really worth to use it ?
    • it should not have good or bad impact on performance
  • Do you have any example of circumstance when you can lose a root volume and, thus promote a LS_Mirror ?
    • it will allow you to continue and serve data from other nodes in case of a failure affecting the hosting aggregate (corruption. raid degradation for example in a case of multi-disk failure, the node freeze (watchdog likely to take a few sec/min to detect, etc... etc...)

Gidi

Gidi Marcus (Linkedin) - Storage and Microsoft technologies consultant - Hydro IT LTD - UK

Re: LS_Mirrors & SVM_DR Best Practices

brasseurm

Thanks Gidi for you reply, very useful.

 

I guess it is not needed neither to snapshot the root volumes?  There's no particular situation where you have to restore it...?

As I know, there are "entry point" for the namespace only and should not contain any user data, but is there any configuration data on it ?

 

I aslo read SVM DR didn't replicate root volumes, but somewhere else I read a contradiction...  which affirmation is correct ?

 

Regarding performance, the LS_Mirrors are entry points for the namespace, is this means IO would go through it, is it worth to place root volumes and LS_mirrors on an faster aggregate ?

 

Thank you guys !

 

 

 
Earn Rewards for Your Review!
GPI Review Banner
All Community Forums
Public