- 论坛徽章:
- 0
|
[原创]AIX操作系统及HACMP群集系统安装步骤
Here is the sample of process:
1/ Log in as root on each node in your cluster.
2/ Ensure that HACMP is not running on any node.. If it is, run smit clstop.
3/ Ensure that you have the same release of HACMP installed on each node in the cluster and that the same fileset components are installed.
# lslpp –L cluster.*
4/ If you find that any filesets are missing for nodes in your cluster, complete the HACMP installation before progressing further.
5/ Ensure that all your nodes have at least 64MB of physical RAM.
# lsattr –El sys0|grep realmem
6/ Ensure that all of your nodes have at least double the amount of paging space in proportion to physical RAM.
# lsps –a
7/ Check to see if each node has the same adapters installed in the same slots.
# lsdev –Cc adapter
8/ If your cluster nodes boot using the Common Desktop Environment, change the system startup mode such that the node boots in ASCII(command line) mode. Otherwise, CDE will hang when IP is enabled.
# smitty dtconfig
choose Command Line
reboot the machine
9/ Check to see if I/O spacing has been enabled on each node.
# smit chgsys
You are suggested to change the ‘High water mark’ to 33 and ‘Low water mark’ to 22.
10/ Check the value of ‘the wall’.
# no –a | grep thewall
11/ Check the physical connection of your IP and non-IP networks. Ensure that each node has at least two network adapters connected to a TCP/IP network and a non-IP connection to the other node in the cluster.
12/ Check the TCP/IP configuration for each node in the cluster.
13/ If there are any errors in the configuration of your network adapters, correct these. Do NOT use the ifconfig command as this only retains the configuration until the next reboot or cfgmgr operation. You can use
# smitty chinet
14/ Check the routing table on each node to ensure that the service and standby adapters have routes to different logical subnets.
# netstat –rn
15/ If there are errors in the configuration of your routes, especially duplicate entries, correct these now. Do NOT use the netstat command as this only retains the configuration until the next reboot or cfgmgr operation. You can use
# smitty mkroute
or
# smitty rmroute
as required.
16/ Check the /etc/hosts and /.rhosts files on each node to ensure that they hold IP labels for the service and standby adapters and server hostnames along with IP addresses.
17/ Check the tty configuration of your serial network. On node 1,
# stty >; /dev/tty1 - command should hang
On node 2,
# stty < /dev/tty1 - command should display on both nodes.
18/ Correct any problems that may exists with the configuration of your serial networks at this stage.
19/ Check that you do NOT have an /etc/netsvc.conf or /etc/resolv.conf file in place on any cluster node at present. If these files are in place, simply rename them to filename.OLD as they may be required later.
20/ Check the network options parameters and ensure that IP forwarding is disabled.
# no –a | grep ipforwarding
if the value is 1,
# no –o ipforwarding = 0
21/ If you are using any form of SCSI technology as your shared disk bus, ensure that no adapter has its external bus ID set to 7. Ideally the adapters should be on SCSI ids 6,5,4 and 3 where possible. i.e.
# lsattr –El scsi0|grep id
22/ If you are using SSA, ensure that you have complete loops between all your cluster nodes and that each node is capable of addressing every disk in all the loops.
23/ Check that the hdisk to PVID numbering of external disks follows the same sequence on each node in the cluster. You may have cross connected SCSI buses or SSA loops.
# lspv
24/ If you find a problem with the cabling of SCSI or SSA shared disks, resolve this before progressing futher.
25/ Check that the number of internal disks matches on each node.
# lsdev –Cc disk
26/ If there is a mismatch between the number of internal disks, perform the following sequence of actions to achieves symmetry:
On the node with the lower number of internal disks:
a/ Unmount any filesystems on non-rootvg volume groups.
b/ Varyoff any non-rootvg volume groups.
c/ Export any non-root volume groups.
d/ List the physical volumes attached to the node(# lspv).
e/ Import your non-root volume groups and change them to not auto-varyon.
f/ Remove all physical volumes that are not part of rootvg (# rmdev –l hdiskN -d).
g/ Create a dummy physical volume using the mkdev command(# mkdev –c disk –t 857MB –s scsi –p scsi0 –w0,5 -d).
h/ Run cfgmgr to force discovery of the removed physical volumes.
i/ List the physical volumes attached to the node(# lspv).
You should now see that external disks have been renumbered as the dummy disk occupies an hdisk location. This extra disk is not shown by lspv because the disk is not in an available state.
27/ Identify all your external volume groups, and which nodes if any have the volume groups varied on(# lsvg –o). Check that your volume groups have the same major number on all nodes and that they are not set to auto-varyon (# chvg –an vgname).
28/ Identify any filesystems that may be present on external volume groups, and their associated LV names.
29/ Download the redbook:
IBM Certification Study Guide: pSeries HACMP for AIX
from http://www.redbooks.ibm.com for further reference. |
|