zero data loss recovery appliance: insider’s ... - oracle · global financial services company...
TRANSCRIPT
Copyright©2018, Oracleand/oritsaffiliates.Allrightsreserved.|
ZeroDataLossRecoveryAppliance:Insider’sGuidetoArchitecture&BestPracticesTimChienDirectorofProductManagementOracleDevelopment
StefanReinersDBAMETRO-nomGmBH
October24,2018
JonySafiSeniorManagerOracleDevelopment
Copyright©2018, Oracleand/oritsaffiliates.Allrightsreserved.|
SafeHarborStatement
Thefollowingisintendedtooutlineourgeneralproductdirection.Itisintendedforinformationpurposesonly,andmaynotbeincorporatedintoanycontract.Itisnotacommitmenttodeliveranymaterial,code,orfunctionality,andshouldnotberelieduponinmakingpurchasingdecisions.Thedevelopment,release,timing, andpricingofanyfeaturesorfunctionalitydescribedforOracle’sproductsmaychangeandremainsatthesolediscretionofOracleCorporation.
Copyright©2018, Oracleand/oritsaffiliates.Allrightsreserved.|
Agenda
Introduction&Architecture
Operational&MonitoringBestPractices
METRONOMCaseStudy
Resources&NextSteps
1
2
3
4
Copyright©2018, Oracleand/oritsaffiliates.Allrightsreserved.|
Agenda
Introduction&Architecture
Operational&MonitoringBestPractices
METRONOMCaseStudy
Resources&NextSteps
1
2
3
4
Copyright©2018, Oracleand/oritsaffiliates.Allrightsreserved.|
TraditionalBackupSolutionsAreNotDesignedforDatabaseTreatDatabasesasJustFilestoPeriodicallyCopy
DailyBackupWindow
Largeperformanceimpactonproduction
DataLossExposure
Losealldatasincelastbackup– novalidation
ManySystemstoManage
Scalebydeployingmorebackupappliances
PoorDatabaseRecoverability
Manyfilesarecopiedbutprotectionstateofdatabaseisunknown
Copyright©2018, Oracleand/oritsaffiliates.Allrightsreserved.|
1010001001101001010101010001010101
0010101010100010101010010101010010100010101010010101000101010100
0011010010101010100010101010010
101000100110100101010101000101
Reliable&CompleteRecovery
EliminateLongBackupWindows
End-to-EndDataProtection
Control&Monitoring
Scale-OutHardware&Storage
ContinuallyValidatesRecoveryStatus
EngineeredDataProtectionForCompleteRecoverabilityoftheOracleDatabase
ZeroDataLossRecoveryAppliance
ArchiveBackuptoCloudStorage
Copyright©2018Oracleand/oritsaffiliates.Allrightsreserved.|
CloudArchive
RemoteReplica
DBBackupCloudService
TapeBackupVirtualFull
DayN
Full
Day0
Incremental
Day1
Incremental
DayN
RecoveryApplianceArchitecture
DeltaPush
CompressedDeltaStore
VirtualFull
ProtectedDBs
ChangedblocksandReal-TimeRedo(nofullbackups)
Datavalidationonreceipt,copy,
restore,periodically
Copyright©2018Oracleand/oritsaffiliates.Allrightsreserved.|
Agenda
Introduction&Architecture
Operational&MonitoringBestPractices
METRONOMCaseStudy
Resources&NextSteps
1
2
3
4
Copyright©2018, Oracleand/oritsaffiliates.Allrightsreserved.|
MAARecommendations
• One1 RecoveryAppliance(RA)systemperdatacenter2
• BackupprimaryandstandbydatabasestotheirrespectivelocalRA
• NoRAreplicationforanydatabasewitharemotestandby
• RestoreoperationcanuseanyRAinanylocation
• DonotmakeanychangestotheRecoveryAppliance
1 ItcanbeaninterconnectedZDLRAconfiguration2 ThedefinitionofaDataCenteralsoincludes“firerooms”
Copyright©2018, Oracleand/oritsaffiliates.Allrightsreserved.|
TapeLibrary
OneWay
Bi-Directional
Hub&Spoke
RemoteDataCenterLocalDataCenter
RAReplica:DataLossProtectionfromSiteDisasters
BENEFITS• ReplicationtoRemote
Applianceprotectsdatafromdisastersorsitefailures
• AutomatedrestorefromLocalApplianceordirectlyfromRemoteAppliance
Copyright©2018, Oracleand/oritsaffiliates.Allrightsreserved.|
LeadingNorthAmericaHealthcareCompany1000+ProtectedDBs,3DataCenters,Bi-Directional+Hub&SpokeReplication
EastCoastDC#2“Hub”(200DBs)
DR NON-PROD
StandbyDBs->LocalRA->DBClone
Standalone->LocalRA->Replicated
MidwestDC(50DBs)
PROD,NON-PROD
Standalone->LocalRA->Replicated->DBClone
EastCoastDC#1(1000DBs)
PROD NON-PROD
StandaloneDBs->LocalRA->Replicated
PrimaryDBs->LocalRA
Replication Replication
Copyright©2018,Oracleand/oritsaffiliates.Allrightsreserved.|
NEW: ZDLRAHighAvailabilityforBackup&Recovery
• Databasebackupandredooperationsautomaticallyfailovertodownstreamreplicaappliance.
• ZDLandpoint-in-timerecoveryoperationsfromdownstreamappliancearefullysupported.
• Virtualfullbackupsonprimaryappliancearesynchronizedtocurrentstatewhenbackonline
PreserveBackup&RecoveryContinuityduringPlannedorUnplannedZDLRAOutages
Copyright©2018, Oracleand/oritsaffiliates.Allrightsreserved.|
RAOsakanormallyreplicatestoRATokyo– allDBsfullyrecoverable
Whenupstreamappliance(RAOsaka)isnotavailable,backupsandredoareredirectedtodownstreamappliance(RATokyo)
• Virtualfullsarecreatedasnormal
Whenupstreamisbackonline,downstreamappliancebackupsaretransferred
• Transferredbackupsareingestedandprocessedintovirtualfulls
• Normalbackupstoupstreamappliancecanberestartedimmediately
Benefits
• PreserveHighAvailabilityduringplannedorunplanneddowntime
• Databasebackup&restore/recoverabilityavailablefromUSorDS
• MAAPresentation andMOSNote2432144.1 NOWAVAILABLE
• www.oracle.com/goto/maa ->ZeroDataLossRecoveryAppliance
X
ZDLRAHighAvailabilityforBackup&Recovery
RAOsaka
Replication
RAOsaka RATokyo
RATokyo
Replication
RAOsaka RATokyo
BackupstoDSAppliance
DSApplianceBackupsTransferredtoUpstream
Copyright©2018, Oracleand/oritsaffiliates.Allrightsreserved.|
BackupFailovertoAlternateAppliance• IncrementalsandRedonormallysenttoPrimaryRA
• AlternateRAservesasbackupstagingareawhenprimaryRAisunavailable,thensyncswithprimaryRAafterwards– Novirtualfullscreatedonalternate,hencerecoverabilitynotsupported
– Spacesizedfor‘n’incrementalsandarchivedlogbackupsduringprimarydowntimeperiod
• Benefits:– Preservesbackupandredoshipmentcontinuityduringplannedmaintenance/upgrades
– PreventslocalFastRecoveryAreasfromfillingupwitharchivedlogs
– Incrementalforeverbackupscontinue
RABostonPrimaryAppliance
XRANYCAlternateAppliance
AlternateAppliancetakesoverbackupsandredotransport
Incrementals&
Redo
“BF_FORWARD”PolicyonAlternateRA:STORE_AND_FORWARD=‘YES’
Copyright©2018, Oracleand/oritsaffiliates.Allrightsreserved.|
GlobalFinancialServicesCompany8000+ProtectedDBs,GlobalDataCenters,BackupFailovertoAlternateAppliance
DB+RAPOD1(600-700DBs)
DBGroup2->RA2->Tape
Failoverto:->RA1
DBGroup1->RA1->Tape
Failoverto:->RA2
DBGroup1 DBGroup2
DB+RAPOD2(600-700DBs)
DBGroup3 DBGroup4
StandbyDBs->LocalRAFailoverto:->RA3
DBGroup3->RA3->Tape
Failoverto:->RA4
EachRAinapodisconfiguredasfailoverfortheother.
• Spaceisreservedfornormal&failoverbackups
StandardizedDBOn-Boarding:
• DataGuardorStandalone?• FailoverConfigRequired?• DBSize+Change%+
Retention->ReservedSpace• ReservedSpaceperiodically
adjustedperspaceneeds
Copyright©2018,Oracleand/oritsaffiliates.Allrightsreserved.|
NEW: ZeroDataLossCross-PlatformDatabaseMigration
• Significantreductionindowntime–lessthan2hourread-onlydowntime,regardlessofDBsize.
• Newautomationtoolsimplifies platformmigrationsteps- especiallyusefulforlargedatabases
• Supportssameandcross-endianplatformmigrationactivities
SimpleDatabaseMigrationtoOn-PremiseorCloud@CustomerusingZDLRA
Copyright©2018, Oracleand/oritsaffiliates.Allrightsreserved.|
SourceDB(11g-12c) DestinationDB(12c)
SOETBS Day1:FullBackup
Export Import
Read-only
MetaData
SOETBS
ZeroDataLossCross-PlatformDatabaseMigrationDramaticallyReduceMigrationTime– FromHours/DaystoFewHoursorLess
• CentralizedRecoveryAppliance“migrationengine” +minimaldowntime(shortread-onlyatend)• Dailyincrementalbackups->virtualfullbackupsonRecoveryAppliance
• Atdestination,restorelatestvirtualfullbackup,priortomigrationwindow• RESTOREFROMPLATFORMXXXFOREIGNDATAFILEYYY
• Whenreadytoswitchover:Atsource,takefinalincrementalandmetadatatablespaceexportinread-only• RECOVERFROMPLATFORMXXXFOREIGNDATAFILECOPYYYYatdestination• IMPORTDataPumpexportfileatdestination
• Destinationopeninread-onlytoverifymigrateddata,thenopenread-writeforbusiness
VirtualFullRestore
Restore&RecoverFinalIncremental
Day2->N:IncrementalBackups
Dayn:FinalIncremental
Read-write
Copyright©2018, Oracleand/oritsaffiliates.Allrightsreserved.|
LeadingGlobalSemiconductorManufacturer
ResultsAchieved• Consolidateandstandardize– Consolidatedseveraldatabaseservers
– CompatiblewithExadata– Multi-Tenantoption
• Reliable,ScalableandHighPerforming
• ImprovedthetimeandcosttobuildandmaintainAnalyticsplatform
• NearzerodowntimemigrationusingZDLRA–RMANDUPLICATE.
• Deliverexceptionalservicetobusinessusers
• Eliminatedfullbackups• ImprovedRTOby4X• Reducedbackupwindowsby2X
ExadataX4-2 ExadataX2-2
CascadeStandby
Incrementalforeverbackups
ActiveDataGuard
ExadataC@C
NewArchitecture
BusinessNeeds• AccelerateGrowth• DriveOperationalExcellence
• CustomerExperience• OperationalEfficiency
• Groworganizationalcapabilities– optimizeinnovation
• Addresscurrent&plannedbusinessgrowthobjectives
SolutionNeeds• Stability• ZeroPreventableOutages• FocusonBusinessOps• IncreaseITagility,self-
serviceandalignmenttobusinessdrivers
ExadataX4-2 ExadataX2-2
ActiveDataGuard
WeeklyFullanddailyincrementalbackup
NASDevice
LegacyArchitecture
Copyright©2018,Oracleand/oritsaffiliates.Allrightsreserved.|
NEW: RASystemActivityReportinBIPublisherTracksRecoveryApplianceActivityandHighlightsAreasofAction
• Displaysinformationforthefollowing:– StateofProtectedDatabases– RASpaceUtilization– StateoftheRecoveryAppliance– Tasks&TaskHistory
Copyright©2018,Oracleand/oritsaffiliates.Allrightsreserved.|
RASystemActivityReport:Healthy?• Itemstowatchfor:Badreport
Watchfor:
IftherearetasksofWORKTYPEinRUNNINGstateandcreatedadayearliertheninvestigate.
IftherearetasksofMAINTENANCEorSBTTYPEpresentandtheircreationtimeisolderthanoneweekthenitshouldbeinvestigated.
AlargenumberoftasksinEXECUTABLEstateforthesameTASK_TYPE
ThesameTASK_TYPEisincreasinginthenumberofjobs(CURRENT_COUNT)overtime.Asystemwithorderingwaitsolderthan1day.
AsystemwithtasksinSTALL_WHEN_WAIT.ThisshouldonlybeseenifOracleSupportistroubleshootingthesystem.
ZDLRA1
Copyright©2018,Oracleand/oritsaffiliates.Allrightsreserved.|
RASystemActivityReport:Healthy?
TheGood:
Minimumcreationtimeforactivetasksiswithinthelast24hoursforworktasks.
Taskhistorystateshouldindicateworkisbeingcompleted.
Reviewothersections
ZDLRA1
Copyright©2018, Oracleand/oritsaffiliates.Allrightsreserved.|
BackupBestPractices• UseTransparentDataEncryption(TDE)insteadofRMANencryption– RMANencryptionwillpreventZDLRAfromcreatingVirtualFullBackups(VB$).
• TDEbackupswillnotgetcompressedontheZDLRA– Encryptedbackupsdon’tgetcompressedbythestorageingeneral– Asizingexercisemustbe(re)-conductedwhenpreparingforTDE– TheincrementalforeverstrategystillappliestoTDEbackups
• UseRAbuilt-incompressioninsteadofRMANcompression– RMANcompressionincursDBCPUutilization&backupsaredecompressed+recompressedonRA
– $rmantarget<targetstring>catalog<catalogstring>backupdevicetypesbtcumulativeincrementallevel1filesperset1sectionsize64gdatabaseplusarchivelognotbackedupfilesperset32;
Copyright©2018, Oracleand/oritsaffiliates.Allrightsreserved.|
UseRecommendedRASoftware• SubscribetoMOSalertsandreferperiodicallytothefollowingnotes:– RecoveryApplianceCriticalIssuesMOSnoteforcriticalissuesalerts (DocID1927928.1)– RecoveryApplianceSupportedVersionsMOSnoteforlatestsoftwareupdate (DocID1927416.1)
• UseRecommendedRASoftwaretoavoidknowncriticalissues– Number1methodtoavoidproblems:Upgradetorecommendedsoftwarerelease– ZeroDataLossRecoveryApplianceSupportedVersions(DocID1927416.1)– ZeroDataLossRecoveryApplianceUpgradeandPatching(DocID2028931.1)– Patchesarecumulativeandinclude
- Bugfixes(Avoidabugbeforeithappens)- Enhancements(Newfeatures,optimizations,etc.)
• CoordinatewithPlatinumPatching– Scheduleearly&OpenProactiveSR
VeryImportant
Copyright©2018, Oracleand/oritsaffiliates.Allrightsreserved.|
Importantthingstoavoid
1. Don’tignore incidents:Resolveandunderstand2. Don’tneglect theRA:Monitorthesystem3. Don’tmake modificationstotheRAconfiguration
i. ZeroDataLossRecoveryAppliance- InstallingThird-PartySoftwareandModifyingInternalApplianceSoftware(DocID2014361.1)
4. Don’t takeperiodicLevel0:VirtualLevel0requiresonlyonelevel0followedbylevel1s5. Don’t backuptoanothermedia:Switchingtoanothermediacanimpactpastandfuture
backupsmadetotheRA
Copyright©2018, Oracleand/oritsaffiliates.Allrightsreserved.|
1. MonitortheRAi. OEMUnifiedManagementDashboard:Reviewtwicedailyandsetupalertsandnotificationsii. RuntheSystemActivityReportdailyandmonitortrends:
ZeroDataLossRecoveryApplianceSystemActivityScript(DocID2275176.1)iii. RunExachkMonthlyandreviewfindings:
HowToupdateexachkoutsideZDLRAInstall,PatchingandUpgrade(DocID2399688.1)iv. ReviewCapacityPlanningReportMonthlyorBi-Monthly
2. UseMultiSection: Setsectionsizeto64GB,thisallowsforefficientprocessinginZDLRA’sflashcachei. Largedatafiles>16TBwillautomaticallyseeanincreaseinsectionsize- sectionsize=sizeof(datafile)/256ii. Smalldatafiles<64GBwillnothavesectionsiii. ForcesFILESPERSETto1
3. Usethelatestlibra (RMANclientsbtlibrarythatsupportsZDLRA):ZDLRA:Wheretodownloadnewsbtlibrary(libra.somodule)(DocID2219812.1)
4. Validatethebackup/restorenetworkfirst:a. UseZeroDataLossRecoveryApplianceNetworkTestThroughputscript(DocID2022086.1) –
requiresqperf,supportedonspecificOSesb. UseHowtomeasurenetworkperformancefromRMANforZDLRAorCloudBackups(DocID2371860.1) –
usesRMAN’s“NETTEST”option,OSagnostic
4 importantthingstodo
Copyright©2018, Oracleand/oritsaffiliates.Allrightsreserved.|
Agenda
Introduction&Architecture
Operational&MonitoringBestPractices
METRONOMCaseStudy
Resources&NextSteps
1
2
3
4
ZDLRA @ METRONOM
10.24.2018
Agenda Introduction
Mission
Best Practices
Target
Summary
1
2
3
4
5
29
Introduction
Introduction
Introduction
Team of 13 DBAs
• ~ 2100 Oracle databases
• AIX, Linux, Windows - Servers
• Database features including:
• RAC
• Data Guard
• GoldenGate for Minimal Downtime Maintenance (MDM)
32
Mission
Mission
• Started POC in 2016
• Replication between two DCs
• Easy Migration
• Go Live April 2017
• Backup of more than 350 productive databases by end of 2017
• Zero Downtime
34
Best Practices
Easy migration• Calculate how much space is needed• Enroll the database to the local ZDLRA• Grant access to the protected database to
a backup user• Modify sqlnet.ora• Create and verify a Wallet• Copy libra.so to the LIB directory• Register database• Set new backup configuration• Enable Block Change Tracking• Run test Backup
Best Practices
addzdlra.sh -d testdb -p gold
MAX_RETENTION_WINDOW
• Set it, but not too aggressive
Don‘t Submit multiple requests to Delete Databases
• A delete can take a lot of time
• If database delete does not progress for some time, contact Support
Don‘t Neglect the RA
• Monitor and adjust the System
• System Activity Report• Clarify Findings with Support
Best Practices
37
Target
Target
• Use ZDLRA to Backup ~900 databases within the next 12 months
• Add more storage to ZDLRA for growth and replication needs
• Implement Minimal Downtime Maintenance patching solution via new RA High Availability for Backup & Recovery procedure (aka Auto-Failover to Replica RA)
• Outsource Backups to Oracle Cloud
RA-based Migration of IBM AIX to ExaCC/OCCATTEND OUR MIGRATION SESSION [TRN4033] TOMORROW, 9 AM, MOSCONE WEST 3004
40
Summary
• Less time for Backups
• Fast restore and recovery
• Easy Migration due to automated processes
• Scalable for data growth
Summary
Thank you foryour attention!
Stefan Reiners
Copyright©2018, Oracleand/oritsaffiliates.Allrightsreserved.|
Agenda
Introduction&Architecture
Operational&MonitoringBestPractices
METRONOMCaseStudy
Resources&NextSteps
1
2
3
4
Copyright©2018, Oracleand/oritsaffiliates.Allrightsreserved.|
Resources&NextSteps
• RecoveryApplianceProductCentral–www.oracle.com/recoveryappliance
• RecoveryApplianceMAABestPracticesCentral–www.oracle.com/goto/maa ->ZeroDataLossRecoveryAppliance
• OpenWorldMAAPresentations–www.oracle.com/goto/maa ->Presentations
• AskTOMBackup&RecoverySessions– asktom.oracle.com->OfficeHours->Searchfor‘backup’
• Getourbusinesscardforfollowupquestions!