Oracle FAQ Your Portal to the Oracle Knowledge Grid
HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US
 

Home -> Community -> Mailing Lists -> Oracle-L -> Oracle8i/NT(cluster) - some web resources, etc.

Oracle8i/NT(cluster) - some web resources, etc.

From: Eric D. Pierce <PierceED_at_csus.edu>
Date: Wed, 25 Jul 2001 18:25:44 -0700
Message-ID: <F001.003555C6.20010725183025@fatcity.com> 






Look like I was wrong, Oracle/NT-cluster installs once via autodiscovery/propogation 
to each node:



http://www.oracle.com/ip/deploy/database/collateral/parallel_server_ds.pdf



(see bottom of page 5)



similar:



http://www.oracle.com/collateral/parallel_server_fo.pdf




http://docs.oracle.com/a91029/0/unixdoc/product_0/index.htm



-



http://docs.oracle.com/DOCS_DOT_PDF_Storage/a73025.pdf


-







more stuff:




http://www.revealnet.com/Products/Oracle/roradm.htm


-



http://www.oreilly.com/catalog/ordbacheck/index.html


-



http://www.geocities.com/tbcox23/




###########



73900.1



Differences between Oracle Failsafe and Oracle Parallel Server on Windows NT



PURPOSE 


To introduce Different options for High Availability 
of Oracle Databases on Windows NT 



SCOPE & APPLICATION 


Intended to give a brief comparison between Oracle Failsafe and 
Oracle Parallel Server. 



Differences between Oracle Failsafe and Oracle Parallel Server on Windows NT 



COST: 


At this time, Oracle Failsafe is a free product if you purchase a database license. 
Oracle Parallel Server requires an additional charge.  Contact your 
Oracle Sales Rep for pricing. 



MSCS versus OSD component from Hardware Vendor: 



Oracle Failsafe requires that Microsoft Cluster Server be installed to manage the 
shared devices in the Cluster.  MSCS must be configured correctly before Failsafe 
will work.  Any platform that is certified to run Microsoft Cluster Server (meaning 
it is on the Microsoft Hardware Compatibility List for MSCS) is also certified 
for Oracle Failsafe. 



Oracle Parallel Server, on the other hand, does NOT require Microsoft Cluster Server. 
Instead, it requires an OSD (Operating System Dependent) component from the 
Hardware Vendor.  OSD on Unix platforms was meant to refer to the hardware and 
operating system because so many hardware vendors supply their own OS as well. 
With Windows NT, the operating system is obviously the same on all platforms. 
However, the hardware vendor still must provide software that has been certified 
to work with Oracle and NT to manage access to the shared drives.  This is what is 
referred to as the Vendor OSD layer.  Each vendor must separately certify their 
hardware platform and OSD component to run with Oracle Parallel Server on 
Windows NT.  A list of certified partners for Oracle Parallel Server on 
Windows NT can be found at the following URL: 



http://www.oracle.com/ip/deploy/database/features/ops/certification/index.html 
  



Shared Nothing Architecture versus Shared Everything Architecture: 



Since Oracle Failsafe requires Microsoft Cluster Server(MSCS), it must adhere to the 
limitations of MSCS, which includes a Shared Nothing Architecture.  What this means is 
that only one node in the cluster can own and/or access the shared drive(s) at any 
given 


time.  Since the database files will be stored on the shared drive, only one node can 
access the database at any given time.  If the owning node fails, control must be 
given 


to the other node before the shared drives and the database can be accessed. 



On the other hand, Oracle Parallel Server utilizes a Shared Everything architecture, 
which 


means that all nodes can access the shared drives (and consequently access the 
database) 


simultaneously.  This gives Oracle Parallel Server the advantage of scalability, as 
the 


workload can be more easily divided among multiple nodes.  Adding to the scalability 
advantage is the fact that there are some platforms certified with Oracle Parallel 
Server 


to have up to 8 Nodes in a cluster.  Due to limitations of Microsoft Cluster Server, 
Oracle 


Failsafe is limited to only two nodes in a cluster.  Windows 2000 Datacenter Server 
supports 


up to four nodes in a cluster. Windows 2000 Advanced Server is limited to two nodes. A 
server 


cluster cannot be made up of nodes running both Windows 2000 Advanced Server and 
Windows 


2000 Datacenter Server. 



NTFS Partitions versus Raw Partitions: 



MSCS and Oracle Failsafe require that shared partitions be formatted using the NTFS 
file system. 


Accessing shared drives requires that both nodes use the same drive letter for each 
shared device. 



Parallel Server, on the other hand, requires that all shared devices to be used in the 
cluster 


remain Raw(unformatted).  This means that you can place only one file on a partition, 
so you will likely have many more partitions in an OPS installation.  You should NOT 
assign drive letters to the partitions.  If NT assigns drive letters, you will likely 
want to unassign them.  To access the RAW partitions, you must use symbolic links, 
rather than drive letters.  To create symbolic links, use the 'Setlinks' utility 
provided by Oracle. 


Symbolic links find the file by using the Device Number, followed by the Partition 
number, 


as seen in Disk Administrator. 



For example:  \Device\Harddisk2\Partition1 would be the first partition on the 3rd 
Hard Disk 


(Start counting from 0) as seen in Disk Administrator.  If you place your control file 
on this 


partition, you will want to use Setlinks to define a symbolic link which acts as a 
shortcut to it. 


Create a .tbl file with the partitions and link names you intend to use.  Below are a 
few lines 


from a sample file called OPSPRIMARY.TBL: 


OPS_CTL01      \Device\Harddisk2\Partition1 
OPS_log1t1       \Device\Harddisk2\Partition2 
OPS_sys01       \Device\Harddisk2\Partition3 
OPS_idx01       \Device\Harddisk2\Partition4 





Run the .TBL file through Setlinks from a DOS Command prompt: 



SETLINKS /F:OPSPRIMARY.TBL 


Symbolic Links are accessed using the convention \\.\<link_name>.  Above, we created a 
symbolic link for the control file on the first partition of the 3rd Device called 
OPS_CTL01. 



You would reference the file in this manner: 



CONTROL_FILES = \\.\OPS_CTL01 


Again, keep in mind that no drive letters are used. 



See the Oracle Parallel Server Getting Started Guide for Windows NT for more details 
on 


the Setlinks utility. 
  



Virtual Server versus Direct connection to a Node: 



Another of the major differences between Oracle Failsafe and Oracle Parallel Server is 
the concept of the 'Virtual Server' used by MSCS and Oracle Failsafe.  The Virtual 
Server 


is meant to be seen as an independent node on the network, regardless of 
which Cluster Node owns it. 



Because of the use of Virtual Servers, an Oracle Failsafe install will require a  
minimum 


of 4 PUBLIC IP Addresses on your network:  one for each node, one for the Cluster, 
and then one for each Failsafe Group created.  The IP addresses used for the Cluster 
and the Failsafe Groups are for the 





Virtual





 machines.  These Virtual machines are 


owned by one or the other node in the cluster, depending on which node is operational, 
and which node you have defined as the preferred node.  However, since these Virtual 
machines have their own IP Addresses assigned, client connect strings will not have 
to change, no matter which Node owns the group. 



When an instance is added to one of the





Virtual





 Failsafe Groups, Oracle Failsafe 


creates services for the instance on each node, It also creates a second Listener 
on each node, which Listens on the Virtual Server IP Address.  The services on each 
node are set to manual, and Failsafe is responsible for starting up the services 
on the owning node.  If a failure should occur, and the secondary node has to take 
over ownership of the group, then all resources in that group will fail over as well. 
This includes the disks, the database, the IP Address and Network Name and the 
Listener. 


The Oracle Intelligent Agent may also be a resource in a group.  Once the Disks and 
the 


IP Address have come online on the secondary node, Failsafe restarts the Group's 
Listener 


and the Services for the instance, along with the database.  A client should then be 
able 


to reconnect with the same connect string, without making any changes to the 
TNSNAMES.ORA. 


With Parallel Server, there is NO Virtual Server.  Each node in the cluster has its 
own 


IP Address, just like any other server on the network.  Since it is a Shared 
Everything 


architecture, each node has its own instance, and services, and all instances can be 
running simultaneously, accessing the SAME database (whose datafiles are on the shared 
RAW partitions).  If one of the instances, or one of the nodes, goes down, than the 
client 


must reconnect to a different instance/node in the cluster.  As mentioned above, all 
instances are still hitting the same database, but reconnecting would require a 
different 


connect string, or modifications to the TNSNAMES.ORA to point to a different 
HOST and/or Instance. 
  
  


###########






         Note:45967.1


 Subject: 

         20 Differences Between Oracle on NT
         and Oracle on Unix
 Type: 
         FAQ
 Status: 
         PUBLISHED

                                       Content Type: 
                                                         TEXT/PLAIN
                                       Creation Date: 
                                                         02-JUL-1997
                                       Last Revision Date: 
                                                         09-APR-2001







 20 DIFFERENCES BETWEEN ORACLE ON WINDOWS NT AND ORACLE ON UNIX






 PURPOSE





 This note enumerates the 20 most obvious differences between
 (Oracle on) Unix and NT.



 SCOPE & APPLICATION





 This note is directed towards dba's and system engineers with
 either a moderate knowledge of (Oracle on) NT or Unix.



 RELATED NOTES





 [NOTE:46001.1] : Oracle and the Windows NT memory architecture
 [NOTE:46053.1] : Windows NT Memory Architecture Overview



...



 4. NUMBER OF DISK DRIVES


 Windows NT is limited to using drive letters A-Z, though use of raw
 partitions can allow disks to be divided up into smaller sections (see
 question 17).



 UNIX has no built-in limitation on number of disk drives.




...




    
  
  1.   FILESYSTEM TYPES AND CAPABILITIES





 Windows NT supports two filesystems - FAT and NTFS. Oracle software and
 datafiles can be installed on either type, with the following provisos:



 Security






 Once a user is connected to an NT server, they must then have
 access to a file to be able to access it.



 NT files can only have a single name (unless using POSIX).
 Files on UNIX can have multiple names via hard or soft links.




    
  
  1.   CLUSTERING





 Windows NT clustering has only become available relatively recently.
 Most of the major hardware vendors support it. Two Oracle products for
 Windows NT clusters are Oracle Parallel Server and Oracle Fail Safe.
 The latter is for two-node clusters, where an instance can only run on
 one node at a given time.



 UNIX clustering has been in existence for several years. Oracle Parallel
 Server has been available on UNIX since early Oracle7 releases. Oracle
 Fail Safe is not available for UNIX.



...



    
  
  1.   PERFORMANCE MONITORING





 Oracle on Windows NT is integrated with the NT Event Viewer and
 Performance Monitor utilities.



 The Event Viewer utility is how an NT administrator views system
 alert messages on NT. Oracle has integrated with Event Viewer such 
 that Oracle startup/shutdown messages and the OS audit trail (if
 you configure OS auditing in Oracle) appear there. Performance
 Monitor is the NT equivalent of the UNIX sar or vmstat command,
 providing detailed resource utilisation data for all processes
 running on the system.



 Oracle has integrated with Performance Monitor such that you can
 view utilisation of operating system and Oracle resources (e.g. file
 write bytes per second - only those related to Oracle, library cache
 hit ratios, etc.). The Oracle Performance Monitor entry in the
 Start Menu > Programs > Oracle for Windows NT starts the standard
 NT Performance Monitor utility but feeds it Oracle-specific data.



 Although multiple instances of Oracle can be run on Windows NT, the
 NT Performance Monitor and Event Viewer utilities can only "see" one
 Oracle instance at a time. For details of how to edit the relevant
 registry variables, see [NOTE:46875.1] or the Getting Started Guide
 for Oracle on Windows NT.



 Oracle on UNIX provides no performance utilities for use at operating
 system level. Utilities such as sar or vmstat must be used to monitor
 Oracle background or shadow processes. These do not provide a graphical
 interface. More sophisticated third-party tools are available on UNIX.




    
  
  1.   RAW PARTITIONS





 Windows NT supports raw (unbuffered) disk partitions, where Oracle can
 store data, log or control files. Each raw partition can be assigned a
 drive letter, but will not be formatted with a filesystem.



 Similarly to UNIX, each raw NT partition will be mapped to a single
 Oracle data, log or control file. Where NT differs from UNIX is
 the naming convention for these files. When referencing a raw
 partition in any Oracle SQL command, the syntax looks like this:



 DATAFILE '\\.\f:' SIZE 49M REUSE



 where f: is the drive letter assigned to the raw partition, referred to
 here as a logical raw file. Windows NT and Oracle also support physical
 raw files, with device names of the form:



 \\.\PhysicalDriveN



 where N is the number of the physical drive, as seen in Disk Administrator.
 Physical raw files would need to be used on a system with more drives than
 available drive letters for the desired number of raw partitions.



 The OCOPY utility can be used to copy data to and from raw partitions,
 in a similar way to the UNIX dd command. Backup of a raw partition must
 be to a filesystem. The NT Backup utility can then be used to copy the
 backup to tape as required.



 On both Windows NT and UNIX, raw partitions must be used for the shared
 data files in a Parallel Server environment, where special Oracle
 utilities are provided for manipulating them.




...



    
  
  1.   HOT BACKUPS





 Oracle on Windows NT supports hot backups using the same backup
 strategy as it on UNIX, i.e. put the tablespaces into backup mode and
 copy the files to the backup location. Then bring the tablespaces out
 of backup mode. By definition, this can all be done while the database
 is up and in use (though it is best to choose a quiet time, when there
 are few transactions).



 The Windows NT feature to be aware of is that NT Backup does not allow
 files in use to be copied, so you must use the OCOPY utility that
 Oracle provides to copy the open database files to another disk location.
 Since OCOPY cannot copy files directly to tape, you will then need to use 
 NT Backup or a similar utility to copy the files to tape, as required.




 20. RELINKING


 Oracle on Windows NT is supplied as a set of executables and dynamic link
 libraries (DLLs). Relinking by the user is not possible on Windows NT, but
 executable images can be modified using the ORASTACK utility, to change
 the size of the stack used by the threads of the Oracle server process.
 This can be useful to avoid running out of virtual memory when using a
 very large SGA, or with thousands of connections. It is recommended that
 this tool should be used under the guidance of Oracle Support.



 On UNIX, object files and archive libraries are linked to generate the
 Oracle executables, and relinking is necessary after operations such as 
 installation of a patch or Net8 protocol adapter.




 REFERENCES


 Oracle for Windows NT - Getting Started




                                                        Oracle Support Services


 .




  Copyright (c) 1995,2000 Oracle Corporation. All Rights Reserved. Legal Notices and
  Terms of Use.



---end---




Note:76438.1


 Subject: 

         Monitoring Oracle Threads using
         Qslice
 Type: 
         BULLETIN
 Status: 
         PUBLISHED

                                       Content Type: 
                                                         TEXT/PLAIN
                                       Creation Date: 
                                                         29-OCT-1999
                                       Last Revision Date: 
                                                         05-APR-2001







 PURPOSE:





 Documents the procedure to use Quick Slice from the NT Resource Kit.  



 Quick Slice is an standalone executable, qslice.exe, included with the NT 4.0
 Resource Kit.  Using this utility, you are able to monitor the cpu usage for 
 the background threads within the Oracle process.  

  


 SCOPE & APPLICATION:





 Oracle analysts and other users who want to monitor Oracle threads.




...



 Step 1:




 You will need to find out the process ids of all the threads running in
 Oracle.  The following SQL will generate this for you:



 select p.spid "Thread ID",

 b.name "Background Process",
 s.username "User Name",
 s.osuser "OS User",
 s.status "STATUS",
 s.sid "Session ID",
 s.serial# "Serial No.",
 s.program "OS Program"




   from v$process p, v$bgprocess b, v$session s
   where s.paddr = p.addr and b.paddr(+) = p.addr;



...




 In this example, Thread Id 114 is PMON, E6 is DBW0, etc.




 Step 2:




 Run Quick Slice on the server.



   If the NT 4.0 Resource Kit is installed, access Quick Slice by selecting:
     Start > Programs > Resource Kit 4.0 > Diagnostics > Quick Slice



 Note:  The Quick Slice executable, qslice.exe, is a standalone application.
 If needed, this executable can be copied from another location and be used
 without the need of any other files.



 Once Quick Slice has been started, a graphical display showing all of the 
 processes on the server will be displayed.  



 To display the individual threads of a process, find the process and
 double click on it.



 For example, to see all of the threads of an Oracle 8.0.x database:





 Another graphical display will appear showing all of the threads and the cpu 
 usage of each. 




 Step 3:




 To identify the threads, match up the TID column to the Thread Id column of 
 the previous SQL statement.  To match the TID numbers with the Oracle Thread process 
ID you will need to 


 convert the TID number from HEX to DEC using the scientific calculator.



 The red bar displays Kernel Time taken by that thread; the blue bar displays 
 user time.


 .




  Copyright (c) 1995,2000 Oracle Corporation. All Rights Reserved. Legal
  Notices and Terms of Use.



---end---




-- 



Please see the official ORACLE-L FAQ: http://www.orafaq.com


-- 



Author: Eric D. Pierce


  INET: PierceED_at_csus.edu


Fat City Network Services    -- (858) 538-5051  FAX: (858) 538-5051
San Diego, California        -- Public Internet access / Mailing Lists


--------------------------------------------------------------------

To REMOVE yourself from this mailing list, send an E-Mail message


to: ListGuru_at_fatcity.com (note EXACT spelling of 'ListGuru') and in
the message BODY, include a line containing: UNSUB ORACLE-L
(or the name of mailing list you want to be removed from).  You may
also send the HELP command for other information (like subscribing).
Received on Wed Jul 25 2001 - 20:25:44 CDT












Original text of this message













  HOME |
  ASK QUESTION |
  ADD INFO |
  SEARCH |
  E-MAIL US