Chapter 2 — Architecture Overview High-Availability Configurations
November 2016 User Guide—Spectra T950 Library
78
Afailovereventoccurswhenthesecondarycontrollerdetectsthatthe
primarycontrollerisnolongercommunicatingonthelibrary’sinternal
communicationbus.Aftersendingaquerytotheprimarycontrollerto
confirmthatitisnolongercommunicating,thesecondarycontrollertakes
overtheexactconfigurationoftheprimarycontrollerandbeginsreceiving
andprocessingthemediachangercommandssentfromthehost.
BlueScaleControllerFailoverallowsbackupstocontinueintheeventthat
librarycommunicationtotheprimarycontrollerisdisrupted.Becauseboth
controllersareconnectedtothesameFibreChannelswitchandmaintaina
synchronizedimageofthelibrary’sinventorydatabase,thefailoveroccurs
withoutmanualintervention.Youcanthentroubleshootthefailed
controlleratyourconvenience.
Afteryouresolvetheproblem,thecontrollerthatoriginallycausedthe
failoverbecomesthesecondarycontrollerinthefailoverpair.Thefailover
pairrevertstotheoriginalprimary/secondaryrelationshipthenexttime
theLCMisrestarted.
TheBlueScaleControllerFailoverfeatureonlyaffectstheoperationofthe
library’scontrollers.ControllerFailoverdoesnotdetectifthecontrolleris
accessiblebythehost.Totakeadvantageofthefeature,thehostmustbe
configuredtoretryfailedcommands.Thisensuresthatcommandsthatfail
whenthelibraryconnectiontotheprimarycontrollerisdisruptedgetre‐
issuedaftertheconnectiontothesecondarycontrollerisestablished.The
retrytimingmustbeconfiguredtoallowforthetimelaginherentinthe
processofdisablingthehost’sconnectiontotheprimarycontrollerand
enablingtheconnectiontothesecondarycontroller.
Exporting Drive Redundancy
Asanalternativetocontrollerfailoverorfordrives,whichcannotbe
configuredwithcontrollerfailover,youcanselectmultipledrivesorRIMs
ascontrollersforasinglepartition,andexportthesamechangerinterface
overthemtoprovideredundancy,aslongasyourdatastoragesoftware
cansupportthis.
Note: Assigningmultipledriveexporterstoonepartitionmayreduce
thetotalnumberofpartitionsthatcanbeconfiguredinthe
library.
Only the secondary controller (acting as primary) is functional after the failover. A
failure on this controller disrupts communications with the hosts. Troubleshoot the
cause of the failover and reconfigure Controller Failover as soon as possible.