source: gpfs_3.1_ker2.6.20/lpp/mmfs/samples/nfscluster/README

Last change on this file was 16, checked in by rock, 17 years ago
File size: 3.1 KB
Line 
1Introduction
2------------
3This directory includes a set of scripts that implements HA-NFS using GPFS
4as the file system.
5
6Network Setup
7-------------
8Setup an IP address for the GPFS cluster, and a separate set of IP addresses
9(one or more) for NFS serving. Use static IP address. DHCP can not be used.
10
11Dependencies
12------------
13(1) Linux heartbeat
14(2) Kernel patch for lockd. (until the code is included in the kernel)
15(3) Fixed sm-notify (until SuSe release the fixed version)
16
17Cluster Setup
18-------------
191. Create a GPFS cluster and file systems.
20
212. Create a small file system for the HA-NFS shared files. This file system
22  should be configured to be mounted when GPFS starts (mmcrfs -A yes).
23
243. Add all file systems but the one for HA-NFS to /etc/exports.
25
264. Stop the GPFS cluster on all nodes with  mmshutdown -a
27   or on one node use         /etc/init.d/gpfs stop
28
295. Modify .sample files appropriately and copy them to /var/mmfs/etc (drop the .sample extension)
30  cp -p /usr/lpp/mmfs/samples/nfscluster/nfsdefs.sample /var/mmfs/etc/nfsdefs
31  cp -p /usr/lpp/mmfs/samples/nfscluster/nfs.nodes.sample /var/mmfs/etc/nfs.nodes
32  cp -p /usr/lpp/mmfs/samples/nfscluster/alert.sample /var/mmfs/etc/alert
33
346. Edit nfsdefs; use the fields as described in the file.
35
367. Edit nfs.nodes; each line is a mapping between GPFS IP address and NFS IP
37   address(es). One line for every node in the NFS cluster.
38
398. Edit alert is optional. This script is invoked to alert an administrator
40   that something is wrong with the NFS cluster. Edit this file to match your
41   needs.
42
439. To complete configuration of GPFS and Linux for HA-NFS on all nodes run the
44   following on one node.
45    /usr/lpp/mmfs/samples/nfscluster/install_ha-nfs.sh -a /var/mmfs/etc/nfsdefs /var/mmfs/etc/nfs.nodes
46
4710. Start the GPFS cluster.
48  mmstartup -a
49 
50 
51Notes
52-----
53Configure GPFS as an HA-NFS cluster:
54 
55  The following option are already set by install_ha-nfs.sh script.
56
57  For better read performance of big files set the following option.
58    mmchconfig nfsprefetchstrategy=1
59
60  To get faster recovery time:
61    Current lease time and wait time are 35 second.
62    Set leaserecoverywait to no less than 10 second.
63      mmchconfig leaserecoverywait=10
64
65    Set leaseDuration to no less than 20 second.
66      mmchconfig leaseDuration=20
67
68user exits:
69  mmfsup, mmfsup.scr, mmfsdown, and mmfsdown.scr are used by HA-NFS
70  all old scripts in /var/mmfs/etc are saved in /var/mmfs/etc.bak
71  After HA-NFS installation the user can add a calls to its exits from
72  mmfsup.scr or mmfsdown.scr
73  The user should NOT create an mmfsup or mmfsdown scripts for now.
74
75bonding:
76  If bonding is used add an entry like the following to ifcfg file
77  for correct monitoring by mii-tool.
78
79  BONDING_MODULE_OPTS='miimon=100 mode=1 use_carrier=0'
80 
81Clean-up:
82  If GPFS or some other component terminated abnormally you can try
83  the following command to avoid reboot. (reboot is recommended)
84    /var/mmfs/etc/cleanup
85
86nfs.nodes example:
87        # The first column is the IP addresses used for the GPFS cluster.
88        # The second column is the IP addresses used to export NFS.
89
90  192.168.2.13            9.1.140.65
91  192.168.2.14            9.1.140.66
92  192.168.1.15            9.1.140.67
Note: See TracBrowser for help on using the repository browser.