roses owe SR ar 50 Sana Bae Azure ASR and SQL Server by Rubén Garrighs | Jun 18, 2018 | Data Pltfom for Developers & DBAS | 0 comments In an on premises environment when we propose solutions to geographical aster, the most common opin is og shipping. The use of asynchronous database ‘miroring 0° availabygroupe with asyrchanousropeas i also common Sut incdes an adstonal ek tha ot usualy contemplated. We ror to the “spo” with tien the changes are ransemed, as quekly asthe network andthe target system alow us. Ths means that when the lester has a human erg, an mporant {ror when we become anare a we wll have ths erorreplcated and apple. Ooviusl,a beter scuton would Beto combine both optans, which re not, fxclusive, wth which we would cover more disaster scanaros ereasing te cost of he elution, ‘Another common alternative in on-premises vial envronments (or physeal ones as wel sto use some loaloluton that alows us to have snapshots of out rmachses replicated goographicaly nthe case of a cloud enronmenn Azure, we can implement tis functionally easly using Azure Site Recovery (ASR), is post, we wil show the operaton of this technology as an alternative fo te use flog shipping tnkng in those eases where we donot wart to nave fo martin the N fog shipping at we need fora SQL Server instance manual or where we wan o conta the cost of he solution even futher For this example, we ar going toc couple of resource groups, ane in the ort of Europe andthe othe inthe west of Europe onitipsaaconindonverazomncanatane sasoso owe oh are 8 Sever pe pa Vreven | © Subscriptions: MMMNMMMMMMMigme 1:0 — 0001 see 2 subscaption? Suitch directovies NAME (0) testasrne (©) tesasewe In tho north European group, wo have deployed a vitual machine with SQL Server 2016 with al he asscciatd resources of a disk, network, se sshogssaconioved ovate ana an aeroses owe SR ar 50 Sana Bae Bitems [_] Show hidden ypes @ testASR_alsk2_770Sdb498179426f09043562b0101632 nae a s —] & tesasrowisk.1.202015790d20429001036073e2bcKe02 BB tcstasia20 Hi eusnip i tates > test aSRNEnet © tesastass 1 Virtual machine Disk Disk Network interface Public iP edsress Storage account Viral nemo Network security group tocaTion North Europe North Europe North Europe North Europe Noth Europe Nosh Europe Nom Europe North Europe In tho west European group, wo wi create a Recovery Services vaul whore the backups and csaste recovery configurations that we perform wilbo stored onitipsaaconindonverazomncanatanesoazot pow are 8 Sera pe test it * Subscription Ss v fesource group (© Create new (@) Use existing testASRWE v * Location West Europe Vv To protect our machine the process is very simple, we just have to goto the properties of he machine and select he "Disaster recovery" option inthe operations category sshogssaconioved ovate ana an 8roses owe SR ar 50 Sana Bae © Auto-shutdown @ backup S Dearne Update management Preve. inventory Preview) change racking Preview) In tho nox settings screen, in our caso, wo wal nly modify tho target resource group, which wo wil be the one we have previously created for to vault: Welcome to Azure Site Recovery "You can replicate your virtual machines to another Azure region for business continuity and cisaster recovery needs. You can conduct periocic DR aril o ensure you meet the ‘compliance needs. The VM will be replicated with the specined settings tothe selected region so that you can recover your applications in the event of outages in sauce region. Leam more about Azure site Recovery. Target region West Europe v Target settinas source ‘arcer vit resource group sean | sw vl] © itn. coeniedoavariacieancarsasiawvar saoan owe SR ar SOL Save Bae vabiny et ot apeeane Theta Viral network seaseNenet [oe ease vetare Storage settings F1Hide cea Cacne storage account | aetna Sone) = SOUNCEMANAGEDOSK ——RFLCAMANAGED OH EDCA MANAGED CTP DPrenMeTLRS TeRASR 2. iw eRASR.dNe-7Pseo4.. [PemmMIAS | Prem teASR Os. (a teRASR OAD pao. [PemiumAS Replication settings Hide tails Mecoeryserieerman [lensed Vauteerouce groups v Repletionptey Camara ay = onitipsaaconindonverazomncanataneross owe SR ar SOL Sava Bae It we look, we can see that we ean configure some optons on his serben that canbe useful For example, we can change the storage type, and use non-premium SOURCE MANAGED DISK REPLICA MANAGED DISK REPLICA MANAGED DISK TYPE (Premium AS} tesisR.dsk2.. new) testash.dsia_77050b4. (ROGGE |e Premium esiSR_OsDi, (pew teasR osdisk 2020. ORO S| ‘Wie can aso contigre the replication poly. By default we have only one ply created, wih 24h of etenton, but we can create a customized one that bes fis our needs 408 inthe OR environment onitipsaaconindonverazomncanataneaaa owe SR ar 50 Sana Bae \When we launch the replestion process, we willbe able to see te process graphically on a map ~ = Source region (North Europe) @ Selected target region (West Europe) © Availabe target regions Wie besr in mind tha this process can take a we depending onthe size ofthe mactine and the confgurtion o be performed. In our case, te configuration took about 10 minutes: onitipsaaconindonverazomncanataneroses owe SR ar 50, Sava Bae ae wars 1 em Enable repieaton © succetstt Protec tem vests Associate repietionpoiey @ Success Fiction potiy 2b-nouretenton poly Associate replicon polity @ succes epicaton poly Amour etenton pay Nap Networks © success Network testane-wetasr Crete protection conaer — @ success ous 2s22a default westopec Nap Neto © sucess Network testare-wet creme aste © success Sener ssea2a-detau-westeuope Crete protatonconsner— @ success lous sera2a-defaut-norhewope crete ate © sucesst sener sera2s-detautenormewope resterepicsion poly Succaeut Repicaton poly 24 nour etenton pot (Once the configurations done, we can see the progress of he raplcation on a global scale onthe machine: Events Replication health Status a 0 oxitipsanconindonverazomncanatann srr Te 352018, 123717 PM 3/6/2018, 235509 62018, 234369 37/2010, 723435 PM 352018, 23404 PM 752018, 234047 52018, 123222 9M sytya0ie, 1232227 sya, 27359 syo/a018, 122731 9M uemon onota7 onora7 ono on000 ono onon3s non onarstroan owe SR ar SOLS Bae Latest Recovery Points Crash-consistent Not available ‘App-consistent Not available In adtion, we wil also have information a the incvidual sk love, wre we would highlight the data of how many pending changes there ate atthe source that has not yet been replicated onitipsaaconindonverazomncanatane soowe oh are 8 Sever pe Operating system disk NAME starus size PENDING DATAIN SOU... REPLICADISK NAME | TARGET DISK NAME R.OsDisk.1.b0... 81% Synchronized 00 68 57 MB UestASR OsDisk.1_b0... testasrosdisk_1_b02. Data disk NAME suas siz PENDING DATAIN SOU... REPLICADISK NAME TARGET DISK NAME 0.00 we test ASR disk2_7705d... testasrdisha_7705dD an the RPO wil be incated, which wil change depending onthe tme elapse since the ast Finally when the replication irihes tha sats wl change "Pate sshogssaconioved ovate ana an seReplication Health Replication health Status RPO Replication Health Replication health Status RPO en owe hare Sa ge Events © Heatthy Protected Events © Heattny Protected a coninieitisacsntsetseetoan owe SR ar 50 Sana Bae From the SQL Server perspecive, by default, backups wll be processed Via VOI: esoe Dae € pe (El e018 1148820 Eade BACKUP DATABASE asecendsy rocteedOpageen 3.17 scene (2000 MB. El Sm2018 114882 alt Bachan BACKUP OATABASEasconedly proceed Opagee 25% seconde 000 MBs). E] W010 14852 AM Backup GACKUP DATABASE ascenddy octet Opogee 31D second (2000). El weamerr4sez all eackap BACKUP DATABASE avceedsy proceed Opageen 299 econde (2000 MB. El sea018 114ss2AM Eackap Daabose backed wp, Daabace: maser cestn cae) 2018/03/05 0649. pages dued 485 tt LN, 228236 128 nt LSM 2283681 mbt ste 14882AM Backup Database bck, Datbaee tetASR.cestondehine 2018/0617 papee duroed 20 fr LSN 34552157 LSM 346901 aber EE Meanie r14es2AM Bachan Database backed up, Database: mode cea tet) 2001/04/08 1338, pagesdurpd: 314 tt LSN 22812 SK. 20896 Frambere El wamtasan ess tnaboe tot ASR. No ver acon eases, [E] nsa018 114282 AM spc62 VO wseresunednasbste maser Nous acn ecu, Ey e018 114892 4M sacs Onan enue on dataate evel No unr acon ere El MeAm18 114822AM Bache Database becker, Database mot can deine) 2716/04/50 46:5 pages Aanped 2138, tt LN 16131238 ok LSN 1613441 rr E] we/018 1140826 mpe59 _Owasrenaned on detmare mas No ar eon erasure Bl weamenasssan sek tase mater Noueractoie mci Howeve# UO rt renuned only you cou cacao bac EF] We2018 TH4B48 AM we63_VOletozanoncatabane tetASR No weet tregured. Hover. LO wnat sane pneu cud ancl bas l_veane 1148<94M _ spc58_VOtozonon catane med No er acon regime However fOr ones pot, you cod cae ebacp. | This has the same implications as when using any system based on tis technology. The main drawback s thatthe backup that Is mado wil validate the diferent chan rom the last backup made by us, so we ry le aunch a dffereial backup we wl find an error as f tere was no valid fll backup. te want ovoid his problem we must make change inthe Windows ris so thet VSS is used by the Azute Backup Agent (hipsidocs microsoft. comin= Uslazuretbackunbackup-azure-ums-roubleshoctttubleshoot-m-snapshotissuesy hos idocs microsoh comfon-aslazure/eackuplbackup-azur-ume- irirocuconsninsomss onitipsaaconindonverazomncanatane sasoszot owe oh are 2 Sever ge ‘Windows VM ‘Azure Backup takes VSS full backups on Windows VMs (read more about VSS full backup). To enable VSS copy backups, the following registry key needs to be set on the VM [HKEY_LOCAL_MACHINE\SOFTHARE\MICROSOFT\BCDRAGENT] " USEVSSCOPYBACKUP"="TRUE” From the point of view of consistancy, we ly have two paints to consi (Ones he “erash-consitent” and the others “app-cansstnt Latest Recovery Points Crash-consistent ‘App-consistent ‘The App-Consistont tho one hat VSS would use bore as we have sad, having a consistent snapshot of tho operation anda he services compatible with VSS insted on te sorvor Howover, with the crash-conssiont, basil wat You gels the same reul tat we would have got “hard rset was apple to a Saver ‘App-Consatent snapshots can be programmed with a manimum frequency of hour We mist bear in mind that we resiore acvash-consistent recovery pln we can have some problems with appleaions tat are not prepared fort Inhe case of SQL Server, shoul not be a problem because the SOL Server database recovery proves fest wil bo able to Mandl hia stuston without problems (emus also thik hat ASR is much more tan 2s" for backups and that DR also offers solitons for VMWare, Hyper-V vieual entonmonts, ee 80 ean be a {00d ayn genera for our OR statagy at the corporate level ructure dhol conioved overseer anéae-n we"mE Seach Cot AIL Properties Lccts BD Automation script @ eadup @ Site Recovery Jobe ® Ales and Events fff) eactup Reporte pouces GE baciup potcies 1B Backup items onitipsanconindoeerazomncanatann prnenmna svesn © Network Mapping & Replication Policies BWM servers © Network Mapping & Replication Policies i Confguration Severs &® Replication Policies 2 Hyperv Stes B Hyper-V Hosts & Replication Policies"GD Replicated items {BE Site Recovery infrastructure: BE Backup infrastructure IB Recovery Plans (Site Recovery) owe SR ar 50 Sana Bae ‘Once he repletion i fished, the decve moment arives, testing tha we can really rage the vitual machine a Western Europe incase ofa crash. For Ns, we are {going to connect tthe anginal machine and we are gong to crete a process that neers eres ina table every mint In ths way, we can very ow teh the fstas once wo restore the DR ervionmert. The flowing sept crealss a database wih a table and insets a record every minute Snaert into toot values (getdare()) ‘We wil obtain the last ecards entered, to have tem asa reference: onitipsaaconindonverazomncanataneroan owe SR ar 50 Sana Bae ISELECT TOP (10) [Fecha) FROM [testASR] [dbo] [test] order by fecha deze m% - FE Resuts gil Messages fecha (pores aaa po1en30e 13484507 20180306 1342.44 893 2oien3es 134144073, 20109306 134044360 20180305 133844 853 701803406 133644833, zotea306 133744823 20100306 133644203, 10 20180395 1335.44 790 ‘The next stp wl be to perform a test aver that basicaly performs the same actions asthe fatover, wil raise the new machines, et, but without affecting the cant replcalion that will conlaue Io work independdy We can select the recovery pol thal we war, the as of which wl be te one wth te lowest RPO: Le Be sevings gf Failover STN se More en a coninieitisacsntsetseet sreesoso owe oh are 2 Sever ge [ iy Failover direction ° ° Recovery Point Choose a recovery point © [ Latest dowest RPO} v se vitual network @ TeStASRNE ne! tis recommended that for atest failover you use a network different fom production network (as specified under Compute and [Network setngs ofthe vitual machine ‘The stops that wil be falowee wil be dopendont on the prot provious ono and ung on. n ths ost, tho fotal me neado environment, but inthis case, thy basicaly consis of crating anew virtual machine clone ofthe has boon less than 4 minates sshogssaconioved ovate ana an wiroses owe SR ar 50 Sana Bae Job nae suns svaRT TE uranion Prerequisites check fr test aover © success ays, 24438"M 0.000 Create test virtual machine @ Successful 3/6/2018, 2:44:38 PM (00:01:35 Prepanng the vtual machine © success 3/0/2018, 246539 TS Start the vitual machine © success yyoo1a,24759M nana Wo il coneetto the machine rated by he et aver: | NAME TPE ‘LOCATION Tid Pecestesvautasache Storage account North rope SS tesase asa Tosdo4nerataetoo0s2se2b01a422-ASRRepia Dik Wes Europe SS testasR ca Trasesorrodaca9043562H01a432-ASRiest ik West rope TT & testasn osdik 1 poabraToadaoA2S0HGOTReRCAE?-ASRReplca Dik eet Europe TT & temtash osDik 1 baneroT90da04290H6073ec48?-ASRtet Die Wert Europe ~ i asrees, Viral macrine West Europe TB tetas esta520b307 6009-296 966 938tcebcoe Network rterace west Europe CT & tesusmaut Recover Services aut West Europe 3 records inserted: ‘And we wil launch the query to check tel onitipsaaconindonverazomncanataneroses owe SR ar 50 Sana Bae SSELECT TOP (10) [Fecha] FROM [testaSR]. [dbo] [test] order by Fecha dese 0% UE Foods gl Neon ‘ec 2018 03-06 134244 899 2018.03.06 1341:44873 201803406 13:40:48 960, 2018-0306 1339:4.653 201803406 133844839 20180346 1337.44 823, 20184346 13:36:44 803, 201803406 13:35:48 750 201803406 13:34:46 780, ‘We can soe that in his caso, we have ony lest one entry 0 the a1 RPO wl have boon greater than 60 seconds, bu less than 120 seconds. Once we have tested that we can I and access the DR machine, we wll complete the process by performing atest cleanup onitipsaaconindonverazomncanatane soroses owe SR ar 50 Sana Bae onittgsaasconiindonyernz mca an aery Service testasrvault cention-palicy eget Windows testasrwe tected dts Standard_D52.v2 Replication Health Replication health status Cleanup test fllover pendi RPO $56 seconds [As on 3/6/20! Notes resto sla comenfomverinciesnean-ahnve Europe Target storage 2c owe oh are 8 Seer pe Bseine geo Stohr CRRA “n testasme-vnet-asr Events 0oan owe SR ar 50 Sana Bae | | __| Testing is complete, Delete test failover Virtual machine( ‘The cloansip procedure i algo ule fas, inthis case, less than § minutos: Job NAME svarus START TIME DURATION Prerequisites check fo test fallver cleanup © successful 3/6/2018, 250:14PM 0010000 Ending test falover © successful 3276/2078, 258:14PM 00.0000 Clean up the test virtual machine © successtl 3/62018, 250-14 PM 00046 In summary we have seen how we can protect a SQL. Server vu machine in Azur rom a disaster by rapleating rom the north region of Europe to Westen Europe The process a esting te DR envionment is saightforwatd ane fast harks Io the automaton sere by Azure. Protecting a single mactune oly 2 mall onitipsaaconindonverazomncanatane aroams owe SR ar 50 Sana Bae stop within 2 DR strategy atthe corporate level, which must take into account many moce factors, as well as the dependencies between the ciferent resources, onitipsaaconindonverazomncanatane 02