CSV and Hyper-V inside cluster
helo all! built 2 nodes cluster hyper-v role.
node a: virtual server 1
node b: virtual server 2, virtual server 3
qustion: happens when csv cordinator node fails? vm in passive host lost hdds?
an example: node csv coordinator role fails. can virtual server 1 immediatly start in node b? can virtual server 2 , virtual server 3 work?
thank you!
simply yes
the storage object maintains registration table contains entry multiple paths available multiple nodes. registration table contains registration , reservation entry paths available nodes. take example of 2 node cluster dual hba , hence see every interface both nodes have register in table unique key each node. rule cannot register , reserve @ same time though there exceptions rule , discuss later. key 8 byte key unique every cluster node. low 32 bits of key contain mask can used identify persistent reservation placed microsoft failover cluster. assuming node 1 hba1 right owning physical disk resource see both registration , reservation entry hba1 interface. in case hba1 fails depending on vendor dsm configuration storage stack automatically start using hba2 without end user intervention or failover of disk resource. every 3 seconds node 1 keeps coming , checking registration table. nodes defend reservations (every 3 secs…but configurable , setting might in troubleshooting issues also) . during split brain scenario challenger node 2 come , enter registration entry in table per rule node 2 cannot register @ same time. needs wait @ least 6 seconds before comes , enters registration entries , own ownership storage object. meanwhile defender node 1 comes after every 3 second and sees registration entries 2<sup>nd</sup> node , scrub entries. challenging node 2 come after 6 seconds not able add reservation entries in table registration entries have been scrubbed node 1. process of successful defense. there may legitimate case node 1 both interfaces unable access storage or node 1 has rebooted or blue screened , in case node 2 should win arbitration of storage object. let’s see happen in case. node 1 hba1 right owning physical disk resource , see both registration , reservation entry hba1 interface. node 1 blue screened, time when node 2 comes register , reserve register , 6 seconds later on revisit put a reservation entry table , own storage object. how storage arbitration , scsi 3 persistent reservation works.
check source more information http://itinfras.blogspot.com/2010/03/storage-architecture-changes-for.html
Windows Server > High Availability (Clustering)
Comments
Post a Comment