Oracle RAC Interview Q&A[1]

7/25/2019 Oracle RAC Interview Q&A[1]

1/14

1. What is the major diference between 10g and 11g RAC?

Well, there is not much diference between 10g and 11gR (1) RAC.

But there is a signicant diference in 11gR.

Prior to 11gR1(10g) RAC, the oowing were managed b! "race CR#

o !atabaseso "nstanceso A##licationso $ode %onitoringo &'ent er'iceso $igh A%aiabiit!

&rom 11gR'(onwards) its cometed $A stac managing and ro%iding the oowing reso*rces as ie the other c*stersotware ie +C# etc.

!atabases

"nstances

A##lications

Cluster %anagement

$ode %anagement

&'ent er'ices

$igh A%aiabiit!

$etwor %anagement (#ro'ides !$*+$*%!$! ser'ices on behal o other traditional ser'ices) and CA$ - ingle Access Client

$aming method, A"/

torage %anagement (with hel# o A% and other new AC lesstem)

2ime snchroni3ation (rather de#ending u#on traditional $2/)

Remo'ed 4 de#endent hang checer etc, manages with own additional monitor #rocess

'. What are "race C*ster Comonents?

Cluster "nterconnect (A"/)

hared torage (4CR*5oting !is)

Clusterware sotware

. What are "race RAC Comonents?

5"/, $ode a##s etc.

-. What are "race erne Comonents (nothing b*t how does "race RAC database difers than /orma singe instancedatabase in terms o inaries and rocess)

Basicall 4racle ernel need to switched on with RAC 4n o#tion when ou con'ert to RAC, that is the diference as it acilitates ew RAC bg#rocess lie 6%4$,6C7,6%!,6% etc.

o t*rn on RAC

8 lin the oracle libraries9 cd 94RAC6&:4%&*rdbms*lib9 mae ; ins:rdbms.m rac:on8 rebuild oracle9 cd 94RAC6&:4%&*bin9 relin oracle

4racle RAC is com#osed o two or more database instances. 2he are com#osed o %emor structures and bacground #rocesses same as thesingle instance database.4racle RAC instances use two #rocesses +&(+lobal &nAtomic Controlle to %emor er'ice (AC%)+2?0;@>+lobal 2ransaction /rocess6%4$>+lobal &n

7/25/2019 Oracle RAC Interview Q&A[1]

2/14

6%!>+lobal &n

7/25/2019 Oracle RAC Interview Q&A[1]

3/14

Cluster nchroni3ation er'ice (C) ocssd,cssd monitor, cssdagent grid owner

&'ent %anager (&5%) e'md, e'mlogger grid owner

Cluster 2ime nchroni3ation er'ice (C2) octssd root

4racle $otication er'ice (4$) ons, eons grid owner

4racle Agent oragent grid owner

4racle Root Agent orarootagent root

+rid $aming er'ice (+$) gnsd root

+rid /lug and /la (+/n/) g#n#d grid owner

%ulticast domain name ser'ice (m!$) mdnsd grid owner

7. What is start* se8*ence in "race 11g RAC? 11g RAC start* se8*ence?

Cic here to now more detais

From this sequence of the log message and timestamp, we get some understanding about

the sequence of clusterware and ASM instance:

1) CSSD and CSSD are up before ASM

!) "otingdis#s used b$ CSSD are disco%ered b$ reading the header of the dis#s, notthrought ASM

&) Startup of C'S ser%ice has to wait until ASM instance is up and the dis#group for

(C' and %otingdis# is mounte

9. As !o* said +oting : "CR ;is resides in A#< ;isgro*s, b*t as er start* se8*ence "C##; starts =rst beore than A#

7/25/2019 Oracle RAC Interview Q&A[1]

4/14

1. Client Connected through CA$ name o the cluster (remember all three "/ addresses round robin resol'es to same ost name(CA$ $ame), here in this case our scan name is cluster01;scan.cluster01.eam#le.com

. 2he re

7/25/2019 Oracle RAC Interview Q&A[1]

5/14

2o add a node, sim#l connect the ser'er to the cluster and allow the cluster to congure the node.

2o mae it ha##en, 4racle uses the #role located in 9+":4%&*g#n#*#roles*#eer*#role.ml which contains the cluster resources, oream#le dis locations o A%. etc.

o this #role will be read local or rom the remote machine when #lugged into cluster and dnamicall added to cluster.

1. What are the =e t!es that A#< s*ort and ee in dis gro*s?

Control les lashbac logs !ata /um# dum# sets

!ata les !B /"6& !ata +uard conguration

2em#orar data les R%A$ bacu# sets Change tracing bitma#s

4nline redo logs R%A$ data le co#ies 4CR les

Archi'e logs 2rans#ort data les A% /"6&

1-. >ist e! bene=ts o A#

7/25/2019 Oracle RAC Interview Q&A[1]

6/14

"s ca#able o listening or all database instances on the same machine in addition to the A% instance

Can run concurrentl with se#arate database listeners or be re#laced b a se#arate database listener

"s named tnslsnr on the 6inu #latorm

12. What is #CA/ istener?

A scan listener is something that additional to node listener which listens the incoming db connection re

7/25/2019 Oracle RAC Interview Q&A[1]

7/14

7/25/2019 Oracle RAC Interview Q&A[1]

8/14

&

7/25/2019 Oracle RAC Interview Q&A[1]

9/14

8 crsctl sto# crs

Assign the networ address to the new networ ada#ters on all nodes using icong=

8icong eth21M.0..1I netmas II.II.II.0 broadcast 1M.0..II

Remo'e the ormer ada#ter*subnet s#ecication and restart Clusterware=

9 oicgdeli ;global eth1*1M.1JN.1.0

8 crsctl start crs

0.

7/25/2019 Oracle RAC Interview Q&A[1]

10/14

ets shell limits i necessar to reoad baancing : how does it wor?

Ko* m*st read hereD here

-0. ;escribe high e%e #tes to con%ert singe instance to RAC?

Read here

-1. What is the diference between A& and &A/ : &C&? at what conditions !o* *se them?

1) A& with tnsnamesa feature of Oracle Net Services for OCI8 clients. TAF is transparent application failover which will move a session to a bacup connection if thesession fails. !ith Oracle "#g $elease 2% &ou can 'e(ne the TAF polic& on the service using 'bms)service pacage. It will onl& wor with OCIclients. It will onl& move the session an' if the parameter is set% it will failover the select statement. For insert% up'ate or 'elete transactions%the application must be TAF aware an' roll bac the transaction. *+S% &ou shoul' enable FCF on &our OCI client when &ou use TAF% it will maethe failover faster.

Note, TAF will not wor with -/C thin.') &A/ with tnsnames with a8 noti=cations tr*e

FAN is a feature of Oracle $AC which stan's for Fast Application Noti(cation. This allows the 'atabase to notif& the client of an& change 0No'eup1'own% instance up1'own% 'atabase up1'own. For integrate' clients% in3ight transactions are interrupte' an' an error message is returne'.Inactive connections are terminate'.FCF is the client feature for Oracle Clients that have integrate' with FAN to provi'e fast failover for connections. Oracle -/C ImplicitConnection Cache% Oracle ata 4rovi'er for .N+T 0O4.N+T an' Oracle Call Interface are all integrate' clients which provi'e the FastConnection Failover feature.)&C&,aong with &A/ when *sing connection oosFCF is a feature of Oracle clients that are integrate' to receive FAN events an' abort in3ight transactions% clean up connections when a 'ownevent is receive' as well as create new connections when a up event is receive'. Tomcat or -/OSS can tae a'vantage of FCF if the Oracleconnection pool is use' un'erneath. This can be either 5C4 05niversal Connection 4ool for -A6A or ICC 0-/C Implicit Connection Cache. 5C4is recommen'e' as ICC will be 'eprecate' in a future release.

-) "/#, with c*sterware either &A/J&C&

ONS is part of the clusterware an' is use' to propagate messages both between no'es an' to application7tiersONS is the foun'ation for FAN upon which is built FCF.

$AC uses FAN to publish con(guration changes an' /A events. Applications can react as those publishe' events in two wa& ,7 b& using ONS api 0&ou nee' to program it7 b& using FCF 0automatic b& using -/C implicit connection cache on the application server&ou can also respon' to FAN event b& using server7si'e callout but this on the server si'e 0as their name suggests it)

Reationshi between &A/J&C&J"/#

4$ -V A$ -V CONS 79 sen'1receive messages on local an' remote no'es.FAN 79 uses ONS to notif& other processes about changes in con(guration of service levelFCF 79 uses FAN information woring with conection pools -A6A an' others.

-'. Can !o* add %oting dis onine? ;o !o* need %oting dis bac*?

7/25/2019 Oracle RAC Interview Q&A[1]

11/14

Fes, as #er documentation, i ou ha'e multi#le 'oting dis ou can add online, but i ou ha'e onl one 'oting dis , b that cluster will bedown as its lost ou @ust need to start crs in eclusi'e mode and add the 'otedis using

crsctl add 'otedis U#athV

-. Ko* ha%e ost "CR dis, what is !o*r ne4t ste?

2he cluster stac will be down due to the act that cssd is unable to maintain the integrit, this is true in 10g, rom 11gR onwards, the crsdstac will be down, the hasd still u# and running. Fou can add the ocr bac b restoring thea*tomatic bac*or im#ort the manual bacu#,

Read comete stes here

--. What haens when ocssd ais, what is node e%iction? how does node e%iction haens? &or a answer wi be same.

Read here

HI. What is 'irtual "/ and how does it worsE

Read here

HJ. !escribe some rac wait e'ents ou e#eriencedE

"race RAC Wait e%ents

and this table,

HL. Can ou modi 5"/ address ater our cluster installationE

7/25/2019 Oracle RAC Interview Q&A[1]

12/14

Fes,read here

HN. ow do ou inter#ret AWR re#ort in RAC instances, what sections in awr re#ort or rac instances are most im#ortantE

Read here.

#date 1;%a;01G, ome #ractical

7/25/2019 Oracle RAC Interview Q&A[1]

13/14

bac$p #atabase ascopy 'or*at 4;new#is$!rop na*e4?

7)RM+>witch#atabase to copy.

:)"eri'y#ba_#ata_'ilesba_te*p_'iles&v9lo! that all 'iles are pointin! to new#is$!rop na*e.

c. Case7Mi!ratin!#is$ !ropto newstora!e bt noa##itional #is$!rop !iven

5)Ta$ethe RM+ bac$p ascopy o' all the #atabases withnew'or*at an#place it inthe #is$.

6)Preparerena*e co**an#s 'ro*v9lo! &v9#ata'ile etc (#yna*ic%eries)

7)Ta$ea bac$p o' p'ile an#*o#i'y the 'ollowin! re'errin! to new#is$!rop na*e

.control_'iles

.#b_create_'ile_#est

.#b_create_online_lo!_#est_5

.#b_create_online_lo!_#est_6

.#b_recovery_'ile_#es

:)stop the #atabase

7/25/2019 Oracle RAC Interview Q&A[1]

14/14

4RAC6&:4%&Q/A2Q94RAC6&:4%&*bin=9/A26!:6"BRARF:/A2Q9XA5&:66/Y=9X4RAC6&:4%&Y*libe#ort 2$:A!%"$ 4RAC6&:4%& /A2 6!:6"BRARF:/A2or "$2A$C& in Zcat *etc*oratab[gre# ;' \]8\[cut ;1 ;d= ;sZdoe#ort 4RAC6&:"!Q9"$2A$C&echo Zsr'ctl status ser'ice ;d 9"$2A$C& ;s 91[ gre# ;i \is running\Zdone

M. !iference between 4A and CR4A is com#lete cluster stac which includes some ernel le'el tass lie managing networ,time snchroni3ation, diss etc, where the CRhas the abilit to manage the resources lie database,listeners,a##lications, etc With both o this 4racle #ro'ides the higha%aiabiit!clustering ser'ices rather onl a^nit to databases.