Shared nothing migration issues (2012)


i'm checking out new features of 2012 , have 2 standalone servers (not clustered) attached own storage (10gb iscsi). i've got dedicated live migration network between 2 servers (10gb). created dummy vm on server #1 few vhds. shared nothing working have questions if setup correct, based on what's going on.

issue #1 - migration seems take long time , isn't using available bandwidth. have 10gb everywhere data moving through systems @ between 800mb - 1.5gb throughout whole process. have great tier1 san (ssds, 15k sas) our sandbox, know source , destination storage aren't stressed. data path looks comes off storagea via iscsi server1, travels on migration network server2, , pushed out storageb via iscsi.

what holdup? why isn't moving faster , using more of 10gb links?  my thought since shared nothing using smb 3.0, maybe smb throttling things? 

issue #2 - possibly related i'm not sure. have multiple vhds on vm (one per lun simplicity).  in perfmon watch migration happen, vhds moving 1 @ time. destination machine writes first lun (with no activity on others), upon completion starts writing next lun, , on.

is limited 1 disk copy @ time? can not parallel copies during shared nothing migration?

i'm checking out new features of 2012 , have 2 standalone servers (not clustered) attached own storage (10gb iscsi). i've got dedicated live migration network between 2 servers (10gb). created dummy vm on server #1 few vhds. shared nothing working have questions if setup correct, based on what's going on.

issue #1 - migration seems take long time , isn't using available bandwidth. have 10gb everywhere data moving through systems @ between 800mb - 1.5gb throughout whole process. have great tier1 san (ssds, 15k sas) our sandbox, know source , destination storage aren't stressed. data path looks comes off storagea via iscsi server1, travels on migration network server2, , pushed out storageb via iscsi.

what holdup? why isn't moving faster , using more of 10gb links?  my thought since shared nothing using smb 3.0, maybe smb throttling things? 

issue #2 - possibly related i'm not sure. have multiple vhds on vm (one per lun simplicity).  in perfmon watch migration happen, vhds moving 1 @ time. destination machine writes first lun (with no activity on others), upon completion starts writing next lun, , on.

is limited 1 disk copy @ time? can not parallel copies during shared nothing migration?

1) shared nothing live migration indeed uses smb 3.0 , it's not possible use available bandwidth smb (at least i've never seen this). can - try enable smb 3.0 related performance optimization features (smb multichannel, smb direct, teaming etc). see:

http://workinghardinit.wordpress.com/tag/shared-nothing-live-migration/

http://msdn.microsoft.com/en-us/library/windows/hardware/jj248719.aspx

http://blogs.technet.com/b/bobh/archive/2013/02/07/technet-radio-hyper-v-local-vs-hyper-v-smb-performance.aspx

http://blogs.technet.com/b/josebda/archive/2013/04/11/technet-radio-series-covers-windows-server-2012-file-server-and-smb-3-0-scenarios.aspx

2) should work. did explicitly enable multiple concurrent vm migration? see:

http://blogs.technet.com/b/canitpro/archive/2013/01/03/shared-nothing-live-migration-goodbye-shared-storage.aspx


starwind iscsi san & nas



Windows Server  >  Hyper-V



Comments

Popular posts from this blog

Edit Group Policy

Hyper-V VM not reaching OS 'Logon' screen

DNS question...