- 论坛徽章:
- 0
|
请教在HP-UX下如何做双机热备
MC/ServiceGuard Installation
MC/ServiceGuard Cluster configuration is based on two nodes machines (namely primary and standby), single LAN connection (data and heartbeat signals), single shared SCSI bus, one shared SCSI data disk as lock disk. Appendix E shows sample MC/ServiceGuard configuration diagram.
Full Installation on MC/ServiceGuard, including cluster and package manager.
command swinstall on node primary and node standby
2 MC/ServiceGuard Configuration
2.1 Basic System Configuration
i Update hosts table (on primary and standby)
Edit file /etc/hosts to include all configured network interfaces of all hosts
ii Setup trusted system (on primary and standby)
Edit file /.rhosts with examples:
primary root
standby root
iii Configure shared volume groups and logical volumes
On primary, create shared volume group(s) and export map file
primary# mkdir /dev/vgshare
primary# mknod /dev/vgshare/group c 64 0x090000
* choose unused minor number eg 0x090000 as necessary
primay# pvcreate -f /dev/rdsk/cXtYd0
* Lock disk controller number X and SCSI address Y
primary# vgcreate /dev/vgshare /dev/dsk/cXtYd0
primary# lvcreate -l 100 /dev/vgshare
primary# newfs -F vxfs /dev/vgshare/rlvol1
primary# mkdir /shared
primary# mount /dev/vgshare/lvol1 /shared
primary# bdf
* checkpoint for filesystem mounting
primary# umount /shared
primary# vgchange -a n /dev/vgshare
primary# vgexport -p -m /tmp/vgshare.map /dev/vgshare
primary# rcp /tmp/vgshare.map standby:/tmp/vgshare.map
On standby, import shared volume group(s)
standby# mkdir /dev/vgshare
standby# mknod /dev/vgshare/group c 64 0x090000
* choose unused minor number eg 0x090000 as necessary
standby# vgimport -m /tmp/vgshare.map /dev/vgshare /dev/dsk/cXtYd0
* Lock disk controller number X and SCSI address Y
** may be different from the ones in primary
standby# vgchange -a y /dev/vgshare
standby# mkdir /shared
standby# mount /dev/vgshare/lvol1 /shared
standby# bdf # checkpoint
standby# vgcfgbackup /dev/vgshare
standby# umount /shared
standby# vgchange -a n /dev/vgshare
On both primary and standby, activate local volume group(s) only
Edit file /etc/lvmrc
a. AUTO_VG_ACTIVATE=1 change to AUTO_VG_ACTIVATE=0
b. under function custom_vg_activation, add local VGs
/sbin/vgchange -a y -s /dev/vg00 # all local VGs
On primary, get ready for MC/ServiceGuard configuration
primary# vgchange -a y /dev/vgshare
2.2 Cluster Configuration
All time intervals, parameters are set as MC/ServiceGuard default. Please notice that the configuration is for functional test purpose only, not a HP supported configuration.
i Query existing hardware & system configuration (on primary)
primary# cd /etc/cmcluster
primary# cmquerycl -v -C /etc/cmcluster/cluster.cfg -n primary -n standby
ii Modify cluster configuration, check and apply (on primary)
primary# vi /etc/cmcluster/cluster.cfg
* refer to appendix A for sample cluster configuration file
primary# cmcheckconf -v -C /etc/cmcluster/cluster.cfg
primary# cmapplyconf -v -C /etc/cmcluster/cluster.cfg
iii Startup cluster (on primary)
primary# cmruncl -v
primary# cmviewcl -v # check point, cluster status is up
iv Auto Startup (on both primary and standby)
primary/standby# vi /etc/rc.config.d/cmcluster
AUTOSTART_CMCLD=0 changed to AUTOSTART_CMCLD=1
2.3 Sample Package and Service Configuration
A sample package (spkg) is configured with single service (ssvr). The service is designed to run a sample service script (sscript, refer to appendix D) which is able to display current time and
hostname on console device every ten seconds.
i Configure package (on primary)
primary# mkdir /etc/cmcluster/spkg
primary# cmmakepkg -p /etc/cmcluster/spkg/spkg.cfg
primary# cmmakepkg -s /etc/cmcluster/spkg/spkg.ctl
primary# chmod 755 /etc/cmcluster/spkg/spkg.ctl
ii Modify package configuration (on primary)
primary# vi /etc/cmcluster/spkg/spkg.cfg
* refer to appendix B for sample package configuration file
primary# vi /etc/cmcluster/spkg/spkg.ctl
* refer to appendix C for sample package control file
iii Duplicate configuration and control files to other sites (on primary)
primary# cd /etc/cmcluster
primary# tar cvf - * | remsh standby '(cd /etc/cmcluster; tar xvpf -)'
primary# cmhaltcl -fv
primary# cmcheckconf -v -C /etc/cmcluster/cluster.cfg -P
/etc/cmcluster/spkg/spkg.cfg
primary# cmapplyconf -v -C /etc/cmcluster/cluster.cfg -P
/etc/cmcluster/spkg/spkg.cfg
iv Startup cluster (on either primary or standby)
primary/standby# cmruncl -v
primary/standby# cmviewcl -v # check point
3 MC/ServiceGuard Installation Acceptance Test
3.1 MC/ServieGuard Package/Service Failure Scenarios
By killing service script, service and causing package fail. The package will be restarted at standby node.
i Startup MC/SG cluster and nodes (on either primary or standby)
primary/standby# cmruncl -v
ii Startup standard package and service on primary
automatically start with cluster, or
primary# cmrunpkg -v
iii Verify cluster, nodes, package and services status on primary
primary# cmviewcl -v
* make sure package auto-switch flag is enable.
iv Terminate service manually on primary
eg by killing process sscript
v Verify cluster, nodes, package and services status on standby
standby# cmviewcl -v
vi Transfer package from standby back to primary manually by shutting down on standby and restarting package on primary
standby# cmhaltpkg -v spkg
standby# cmmodpkg -e -n primary spkg
standby# cmrunpkg -v -n primary spkg
standby# cmmodpkg -e spkg |
|
免费注册
发表于 2004-07-13 20:43
