Você está na página 1de 31

2006 BEA Systems, Inc.

257
Multiple Machine (MP Model) Domain
Configurations
At the end of this module, you will be able to:
Understand the components of the MP model
architecture
Configure a listener process for each remote machine
Configure and deploy a distributed Tuxedo
application
Perform MASTER machine failover administration
Module 9
Multiple Machine (MP Model) Domain Configurations-1
2006 BEA Systems, Inc. 258
Road Map
1. Basic MP Model Configuration
Distributed Architecture
Managing the tlisten Process
Configuring Basic NETWORK Section Settings
Providing for a Backup MASTER
2. Additional Configuration and Administration
Multiple Machine (MP Model) Domain Configurations-2
2006 BEA Systems, Inc. 259
Distributed Application Benefits
Servers can be configured to execute within
separate groups and machines.
Benefits of distributed applications include:
Scalability
Reliability
Maintainability
Modularity
Multiple Machine (MP Model) Domain Configurations-3
2006 BEA Systems, Inc. 260
A Distributed Tuxedo Application
N
e
t
w
o
r
k

Master
DBBL
BBL
BB Bridge
User
Application
Server
User
Application
Server
User
Application
Server
Admin
Server
tlisten
TUXCONFIG
Remote #1
Bridge
BBL
BB
User
Application
Server
User
Application
Server
User
Application
Server
Admin
Server
tlisten
TUXCONFIG
(copy)
.
.
.
tlisten
Remote #n
TUXCONFIG
(copy)
Master
Multiple Machine (MP Model) Domain Configurations-4
2006 BEA Systems, Inc. 261
Multiple Machine Boot Sequence
Master
DBBL
BBL
BB
Bridge
User
Application
Servers
System
Servers
TUXCONFIG
tmboot
Non-Master
BBL
BB
Bridge
User
Application
Servers
System
Servers
TUXCONFIG
(copy)
tlisten
Bootstrap
Bridge
1
10
2
3
4
7
8
9
5
6
5
6
11
N
e
t
w
o
r
k

Multiple Machine (MP Model) Domain Configurations-5
2006 BEA Systems, Inc. 262
Starting a Listener at all Sites
tlisten is required to be started on all remote sites
Recommended on master site
Usually started at O/S boot time with other daemons
Start-up scripts in UNIX; installed as a Windows service
UNIX: UID must be set since tlisten will not run as root
TUXDIR must be set in the environment
Must exactly match NLSADDR in NETWORK section


Syntax for starting tlisten:
> tlisten [-d device] l nlsaddr [-u uid][-z bits][-Z bits]
UNIX example:
> tlisten l //192.34.110.5:3051 u 101 z 128
Windows example:
> tlisten l //dnsname:3051 Z 0
Multiple Machine (MP Model) Domain Configurations-6
2006 BEA Systems, Inc. 263
BEA Administration Program
In Windows, the tlisten process can also be
configured from the BEA Administration Program.
At installation, port 3050 is configured and started by default
Choose Start > Settings > Control Panel then double-click
the icon.
You can also manage environment variables and IPC
Resource settings from this program.

Not available in Version 6.5
Multiple Machine (MP Model) Domain Configurations-7
2006 BEA Systems, Inc. 264
Configuration Parameter Options
Networked configurations: UBBCONFIG Parameters
A valid network group name (Default: DEFAULTNET) NETGROUP
The device name used by BRIDGE process to access network BRIDGE
The network listening address of the tlisten process NLSADDR
The network listening address of the BRIDGE process NADDR
NETWORK
String value used to compare against all other machines TYPE
MACHINES

Required
RESOURCES
Set to MP (Enables all other networking parameters) MODEL
BBLQUERY timeout (Multiplier of SCANUNIT-Default:20s) DBBLWAIT
DBBL checking BBLs period (Multiplier of SCANUNIT-Default:5m) BBLQUERY
Set to LAN OPTIONS
Description SECTION/PARAMETER
MINENCRYPTBITS
MAXENCRYPTBITS
Min & Max encryption levels: (0 | 56 | 128) bit key size



Multiple Machine (MP Model) Domain Configurations-8
2006 BEA Systems, Inc. 265
MP Mode Configuration Example
10
0101
1110
Example UBBCONFIG text file:
*RESOURCES
IPCKEY 39211
MASTER SITE1
MODEL MP
OPTIONS LAN
*MACHINES
gumby LMID=SITE1 TYPE=SUNSPARC
TUXCONFIG=/home/apps/tuxconfig
APPDIR=/home/apps TUXDIR=/usr/bea/tuxedo9.0
SNOOPY LMID=SITE2 TYPE=HPINTEL
TUXCONFIG=d:\apps\tuxconfig
APPDIR=d:\apps TUXDIR=c:\bea\tuxedo9.0
*GROUPS
APPGRP1 LMID=SITE1 GRPNO=1
APPGRP2 LMID=SITE2 GRPNO=2
*NETWORK
SITE1 NADDR=//gumby:4553 NLSADDR=//gumby:3050
SITE2 NADDR=//snoopy:8726 NLSADDR=//snoopy:3050
*SERVERS
. . .
The machines LMIDs are used as the
Identifiers in the NETWORK section
By default (if TYPE is not
specified) Tuxedo assumes that
all machines are the same
Multiple Machine (MP Model) Domain Configurations-9
2006 BEA Systems, Inc. 266
Failure Scenarios
N
e
t
w
o
r
k

Master
DBBL
BBL
BB Bridge
User
Application
Server
User
Application
Servers
User
Application
Server
System
Servers
tlisten
TUXCONFIG
Remote
Bridge
BBL
BB
User
Application
Server
User
Application
Servers
User
Application
Server
System
Servers
tlisten
TUXCONFIG
(copy)
Master
But what happens if the Master machine fails ...
Each BBL monitors all servers on its machine, restarting failed
servers if they are configured for restart.
The DBBL monitors all BBLs, restarting a BBL after failure.
The Master BBL monitors the DBBL, restarting it after failure.
Multiple Machine (MP Model) Domain Configurations-10
2006 BEA Systems, Inc. 267
Backing up the MASTER machine
The MASTER machine is critical for full operations
Administrative changes can be done only on the MASTER
DBBL polls other machines to check whether they are still
running; also communicates BB changes to other BBs
For multiple machine (MP model) configuration,
specify a backup for the MASTER

*RESOURCES
MODEL MP
OPTIONS LAN
MASTER
...

SITE1 is the MASTER
SITE2 is the backup
SITE1,SITE2
UBBCONFIG
Multiple Machine (MP Model) Domain Configurations-11
2006 BEA Systems, Inc. 268
Transferring MASTER Role to BACKUP
Switching the MASTER role to the backup enables
administration operations to be uninterrupted if the
original MASTER machine fails for any reason.
Use the master (m) subcommand of tmadmin to
switch the MASTER role to the BACKUP machine
when:
The MASTER node needs to be shut down for maintenance
The MASTER node is no longer accessible
Only the designated backup machine may be used to
run the master command.
Multiple Machine (MP Model) Domain Configurations-12
2006 BEA Systems, Inc. 269
tmadmin: master sub-command
To switch the MASTER role to a BACKUP:



When the MASTER node is accessible to the BACKUP node
$ tmadmin
. . .
> master
are you sure? [y,n] y
Migrating active DBBL from SITE1 to SITE2, please
wait...
DBBL has been migrated from SITE1 to SITE2
> quit
1 Run tmadmin from the BACKUP machine
2 Use the master (m) subcommand
Multiple Machine (MP Model) Domain Configurations-13
2006 BEA Systems, Inc. 270
Restoring MASTER from the BACKUP
To restore the MASTER role after a failure of the
original Master machine:






NLSADDR must have been configured for the Master

1
Repair Master node if necesary; restart
tlisten
2 On the BACKUP machine (now Master), run tmboot -B
SITE1 to re-start the Tuxedo administrative servers
3 Run tmadmin from the original MASTER node
4 Use the
master(m)
command
5
Boot the application servers if necessary with
boot -l
SITE1
Multiple Machine (MP Model) Domain Configurations-14
2006 BEA Systems, Inc. 271
Section Review
Deploy the tlisten process on Tuxedo machines
Configure the NETWORK section to support distributed
applications
Configure and perform MASTER machine backup
In this section, we learned how to:
Multiple Machine (MP Model) Domain Configurations-15
2006 BEA Systems, Inc. 272
Lab Exercise
For details on the exercise, refer to the Lab Guide.
If questions arise, ask the instructor.
The instructor will determine the stop time.
Lab 10 MMC: Configure and Deploy a
Multiple Machine Tuxedo Application
Multiple Machine (MP Model) Domain Configurations-16
2006 BEA Systems, Inc. 273
Road Map
1. Basic MP Model Configuration
2. Additional Configuration and Administration
Enhancing BRIDGE Functionality
Configuring the NETGROUPS Section
Detecting Partitioned Machines
Load Balancing Network Requests
Multiple Machine (MP Model) Domain Configurations-17
2006 BEA Systems, Inc. 274
Multiple BRIDGE Connections
The BRIDGE process can be configured for multi-
threaded execution.
Set BRTHREADS=Y in the MACHINES section
The BRIDGE process can also utilize multiple network
channels (available in all versions).
Multiple network connections between machines can
be defined and prioritized in the NETGROUPS section.
Multiple connections can be used to:
Increase reliability: failover to an alternate connection; fail-
back when the original connection is restored
Increase bandwidth: connections can be used in parallel to
increase network throughput
Versions 8.1 +only
Multiple Machine (MP Model) Domain Configurations-18
2006 BEA Systems, Inc. 275
The NETGROUPS Section
The NETGROUPS section allows for the configuration
of network groups referenced in the corresponding
NETWORK section.
X
Required
NETGRPNO
Unique network group number
NETPRIO
Priority of the group [08192, default is 100]
Description Parameter
Syntax of NETGROUPS Section:
*NETGROUPS
netgroup_name NETGRPNO=number NETPRIO=number

Multiple Machine (MP Model) Domain Configurations-19
2006 BEA Systems, Inc. 276
Example Network Group Configuration
In this NETGROUPS example, AB traffic will flow in
parallel across the Magenta and Green circuits.
Example NETGROUPS Section Configuration:
*NETGROUPS




BLUENET NETGRPNO = 9 NETPRIO = 200
GREENNET NETGRPNO = 13 NETPRIO = 300
DEFAULTNET NETGRPNO = 0 NETPRIO = 100
MAGENTANET NETGRPNO = 125 NETPRIO = 300
DEFAULTNET
BLUENET
MAGENTANET
GREENNET
A B C D E
Multiple Machine (MP Model) Domain Configurations-20
2006 BEA Systems, Inc. 277
Example Network Configuration
Example NETWORK Section Configuration:











NLSADDR=...
NLSADDR=...
NLSADDR=...
NLSADDR=...
NLSADDR=...
NADDR="//E_BLUE:5726" SiteE NETGROUP=BLUENET
NADDR="//D_BLUE:5726" SiteD NETGROUP=BLUENET
NADDR="//B_GREEN:5725" SiteB NETGROUP=GREENNET
NADDR="//C_CORPORATE:5723" SiteC NETGROUP=DEFAULTNET
NADDR="//D_CORPORATE:5723" SiteD NETGROUP=DEFAULTNET
NADDR="//A_CORPORATE:5723" SiteA NETGROUP=DEFAULTNET
NADDR="//B_CORPORATE:5723" SiteB NETGROUP=DEFAULTNET
NADDR="//A_MAGENTA:5724" SiteA NETGROUP=MAGENTANET
NADDR="//B_MAGENTA:5724" SiteB NETGROUP=MAGENTANET
NADDR="//E_CORPORATE:5723" SiteE NETGROUP=DEFAULTNET
NADDR="//A_GREEN:5725" SiteA NETGROUP=GREENNET
NADDR="//C_MAGENTA:5724" SiteC NETGROUP=MAGENTANET
A B C D E
DEFAULTNET
BLUENET
MAGENTANET
GREENNET
Multiple Machine (MP Model) Domain Configurations-21
2006 BEA Systems, Inc. 278
Network Group Priority Example
In the example priority settings below, connections to
machine C are preferred through the Magenta Group.
The DEFAULTNET group should always have the
lowest priority
DEFAULTNET, NETPRIO=100
BLUENET, NETPRIO=200
MAGENTANET, NETPRIO=300
GREENNET, NETPRIO=200
A B C D E
Multiple Machine (MP Model) Domain Configurations-22
2006 BEA Systems, Inc. 279
Network Partitions
A network partition refers to a machine that cannot
access the master machine.
Network partitions can be caused by:
Transient network failures
Severe network failures
Master machine failure
Non-master machine failure
Bridge process failure
Multiple Machine (MP Model) Domain Configurations-23
2006 BEA Systems, Inc. 280
Detecting Partitioned Networks
To detect network problems, use:
ULOG to check for network error messages.
tmadmin to gather information about network.
tmadmin provides these subcommands:
Prints whether machine is partitioned. If not,
shows machines connected to and messages
in and out
printnetwork (pnw)
[machineList]
Prints information about a server. Useful to
find out if a server in a particular machine is
up and running and if it is processing
requests in its queue
printserver (psr)
[-m machine ]
[-g groupname ]
[-i srvid ]
[-q qaddress ]
Description Command
Multiple Machine (MP Model) Domain Configurations-24
2006 BEA Systems, Inc. 281
Troubleshooting Example
Example use of the printnetwork(pnw) and printserver(psr) commands:

$ tmadmin
> pnw SITE2
Could not retrieve status from SITE2

> psr -m SITE1
a.out Name Queue Name Grp Name ID Rq Done Load Done Current Service
---------- ---------- -------- -- ------- --------- ---------------
BBL 30002.0000 SITE1 0 - - ( - )
DBBL 123456 SITE1 0 121 6050 MASTERBB
simpserv 00001.0001 GROUP1 1 - - ( - )
BRIDGE 16900672 SITE1 0 - - ( DEAD )
Multiple Machine (MP Model) Domain Configurations-25
2006 BEA Systems, Inc. 282
Restoring Network Connections
The BRIDGE process attempts to recover
automatically (reestablish the connection) after a
network failure.
You can explicitly recover from a network outage by
using the tmadmin reconnect subcommand:
rco Syntax: rco partitionedLMID nonPartionedLMID
rco Example: rco SITE3 SITE1
Multiple Machine (MP Model) Domain Configurations-26
2006 BEA Systems, Inc. 283
Partitioned Machine - Clean Up
A partitioned machine is one that has lost contact with
the Master machine, either due to a network or a system
failure.
If the machine specified is partitioned, pclean
removes all machine entries from all non-partitioned
Bulletin Boards.
If machine specified is not partitioned, pclean simply
executes a bbclean on the indicated machine.
pclean Syntax: pcl partitioned_machine_LMID
pclean Example: pcl SITE3
Multiple Machine (MP Model) Domain Configurations-27
2006 BEA Systems, Inc. 284
Load Balancing Network Requests
Load at remote sites can be tuned with:
TMNETLOAD environment variable
NETLOAD parameter in MACHINES section
Load balancing behavior occurs when services are
available both locally and remotely.
They are added to LOAD factor of remote connections;
favoring local connections
Remote Tuxedo Machine
*SERVERS
Server1
Server2
Load=LOAD
+ NETLOAD
Local Tuxedo Machine
*SERVERS
Server1
Server2
Load=LOAD
If NETLOAD or TMNETLOAD is configured
CLIENT
Multiple Machine (MP Model) Domain Configurations-28
2006 BEA Systems, Inc. 285
NETLOAD Example
...
*MACHINES
goofy LMID=SITE1
...
pluto LMID=SITE2
NETLOAD=450
...
*GROUPS
APPGRP1 LMID=SITE1 GRPNO=1
APPGRP2 LMID=SITE2 GRPNO=2

*SERVERS
audit_server SRVGRP=GRP1 SRVID=10 MIN=5
audit_server SRVGRP=GRP2 SRVID=20 MIN=5
*SERVICES
Audit LOAD=50
UBBCONFIG
Assuming that client requests originate on
goofy, this configuration will cause ten times as
many Audit requests to stay on goofy than
will be sent to servers on machine pluto
Multiple Machine (MP Model) Domain Configurations-29
2006 BEA Systems, Inc. 286
Section Review
Configure the BRIDGE for multiple network
connections
Identify network partitioning problems
Load balance network requests
In this section, we learned how to:
Multiple Machine (MP Model) Domain Configurations-30
2006 BEA Systems, Inc. 287
Module Review
Configure the components of the MP model
architecture
Configure a listener process for each remote machine
Configure and deploy a distributed Tuxedo application
Perform MASTER machine failover administration
In this module, we learned how to:
Multiple Machine (MP Model) Domain Configurations-31

Você também pode gostar