From stephane.paquette@standardlife.ca Mon, 07 Jul 2003 13:37:12 -0700 From: "Stephane Paquette" Date: Mon, 07 Jul 2003 13:37:12 -0700 Subject: RE: service name, sid .. Message-ID: MIME-Version: 1.0 Content-Type: text/plain ... when the fecal matter hits the rotary oscillator.   Politically correct ????       Stephane Paquette Administrateur de bases de donnees Database Administrator Standard Life www.standardlife.ca Tel. (514) 499-7999 7470 and (514) 925-7187 [EMAIL PROTECTED]   -----Original Message-----From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]On Behalf Of Goulet, DickSent: Monday, July 07, 2003 4:00 PMTo: Multiple recipients of list ORACLE-LSubject: RE: service name, sid . AK,       You can make it as simple or complicated as you want.  Personally I prefer the KISS principle.  It's easier to troubleshoot when the fecal matter hits the rotary oscillator.   Dick GouletSenior Oracle DBAOracle Certified 8i DBA -----Original Message-----From: AK [mailto:[EMAIL PROTECTED]Sent: Monday, July 07, 2003 3:34 PMTo: Multiple recipients of list ORACLE-LSubject: Re: service name, sid .. Tanel,   Thanks for elaborate information , I really liked it . Still why do u need to provide db name in tnsnames, all one should know is which instance to connect . Instance know , which db to connect . why do net8 bother about dbname ?   -ak     ----- Original Message -----
From: Tanel Poder To: Multiple recipients of list ORACLE-L Sent: Thursday, July 03, 2003 10:49 AM Subject: Re: service name, sid . Hi!   There's actually more:   db_name - identifies database name, has to be the same what is stated in controlfile (using create database or create controlfile). You can have several databases with same db_name in one server, there is no restriction.   SID - When we actually want to open and use the database, we have to start an instance which will be servicing the database. When starting instance, whe have to specify SID (system identifier) for it. This is actually just a operating system name for instance (or should I say SGA shared memory segments). As you know, SID is specified using ORACLE_SID OS environment variable and is only used by listener, when spawning new processes or when attaching directly to SGA using bequeath protocol. You can not have instances with same SID in one server, even if you use different oracle homes.   instance_name - An Oracle parameter for specifying Oracle instance name. Seems that operating system doesn't know anything about it. For example, if you use SID_LIST parameters in your listener.ora, then listener always knows how to spawn processes for given SID, because (almost) all it takes to start another server process, is the location of oracle executable and SID value. But if you want to connect using instance name (not SID itself) then Oracle instance actually has to register itself with listener before any server processes can be spawned. Note that ORACLE_SID and instance_name variables do not have to match. (Tested on 9.2.0.1 on W2K). If my SID is ORCL for example and instance_name is TEST, then I can connect using both SID=ORCL and SID=TEST in CONNECT_DATA section of TNSNAMES (again, the instance name has to be registered with listener). V$INSTANCE still shows SID (ORCL) in instance name fielt.   As we know, in OPS and RAC environment you could have several instances servicing one database, this is one of common reasons why db_name and instance_name would be different. (ORCL for db_name, ORC0, ORC1, ORC2, etc.. for instance names for example). Btw, does anyone know if I can have the same instance name in all RAC nodes? RAC shouldn't care much, because it uses instance numbers anyway?   service_names - this is an additional layer for logically spreading work across nodes in RAC (and possibly other load balanced environments such replicated databases and even logical standby). For example, if you have 4 node RAC and have defined two "services" such OLTP and REPORTING, then you could assign OLTP for nodes and REPORTING for one, but during month end reporting you could still keep OLTP on 3 nodes, 1 node purely for REPORTING and add REPORTING to one or more OLTPs as well. So, when user's are connecting, they don't request a specific node or instance, they just request a service. And in load balancing environments then users are directed to instances, where appropriate service is defined. It gives somewhat flexibility, but I've never used it myself (thus my explanation here might wrong ;)   Cheers, Tanel.
----- Original Message -----