discuss: Redhat AS2.1 HA Cluster miniHOWTO
Subject:
Redhat AS2.1 HA Cluster miniHOWTO
From:
####@####.####
Date:
15 Apr 2003 18:23:41 -0000
Message-Id: <7A4BB7EC40E7D5118B55009027A19B550281E229@CWDXMEDC01>
Attached is an outline for how I would like to write this. FEEDBACK
PLEASE!!!
Regards,
Derek Black
Storage Analyst
Dell Computer Corp.
EMEA Enterprise Expert Centre
+353 (0)1 279 2027
1) Install RedHat Advanced Server 2.1 on both cluster nodes.
2) Set up appropriate peripherals, shared storage, etc. Ensure to connect a
Cat5 Xover cable on to the second network interface of each machine in order
to establish heartbeat functionality.
3) Configure IP addresses/host information:
a. Node1
i. Hostname
ii. Eth0 IP address (primary channel)
iii. Eth1 IP address (heartbeat channel)
iv. Cluster IP address
v. Service IP address and alias (as required)
vi. Edit /etc/hosts to reflect these mappings
b. Node2
i. Hostname
ii. Eth0 IP address (primary data channel)
iii. Eth1 IP address (heartbeat channel)
iv. Cluster IP address
v. Service IP address and alias (to match the settings on Node1)
vi. Edit /etc/hosts to reflect these mappings
4) Partition the shared storage thus:
a. 2 Quorum partitions: minimum of 10MB, preferably on the same disk or in
same RAID set and separate to frequently-accessed data. Edit the
/etc/sysconfig/rawdevices file to bind quorum partitions to raw devices. A
sample entry in the file would look like this:
/dev/raw/raw1 /dev/sdb1
/dev/raw/raw2 /dev/sdb2
b. This configuration must be performed on both nodes.
c. Partition other storage as necessary.
5) Configure Cluster Manager Software
a. Cluster Manager RPM is installed as a part of the default installation of
RedHat Advanced Server 2.1. This can be verified by querying the clumanager
package.
b. Ensure the two quorum partitions are accessible by running cludiskutil
-t.
c. Run cludiskutil -p on both nodes. This prints header data structure info
for the quorum partitions. Ensure the output is exactly the same on both
nodes to make sure the nodes are accessing the same quorum partitions. The
output for this command should look similar to the following:
----- Shared State Header ------
Magic# = 0x39119fcd
Version = 1
Updated on Thu Sep 14 05:43:18 2000
Updated by node 0
--------------------------------
d. Run /sbin/cluconfig utility. This edits the /etc/cluster.conf file
e. The cluconfig utility will prompt for information as set up in steps 3.a
and 3.b. Specifically, it will prompt for the following information:
i. Cluster name
ii. IP address for cluster alias
iii. Name of cluster member [node]
iv. Number of heartbeat channels (minimum of 1)
v. Heartbeat channel type (serial or network)
vi. Hostname of node on heartbeat channel0
vii. Primary quorum partition (refer to raw_device1)
viii. Shadow quorum partition (refer to raw_device2)
ix. Power switch information (not necessary, but recommended)
x. Repeats questions for node2
xi. Displays config summary and prompts to confirm writing to file. Also
asks if the cluster can be administered locally or remotely through a GUI.
f. After installing on one host, DO NOT run this configuration utility on
the other host, otherwise the cluster will have 2 primary nodes and get
confused. If this happens, when the cluster service on one node stops, it
will reboot the other node.
g. Download the cluster configuration settings to the secondary node by
invoking the script: cluconfig --init=raw_device (e.g. cluconfig
--init=/dev/raw/raw1)
h. Restart the cluster service: /etc/rc.d/init.d/ on both nodes.
6) Configuring failover services - points to remember
a. Allocated partitions should not appear in /etc/fstab. Clustering software
will manage this.
b. Services should be stopped and disabled in /etc/rc.d/init.d (chkconfig
-del script_name). Clustering service will manage this.
c. Assign an IP to the service. Modify /etc/hosts to reflect this.
d. For highly available applications, it is recommended to create a mount
point and filesystem on the shared storage for the service.
e. Individual services must be configured on both nodes in order to fail
over.
f. To configure a service, launch the cluster administration utility,
cluadmin.
g. At the cluadmin prompt (cluadmin>), services may be added by typing
service add service_name. The command will prompt the user for the following
information about configuring a service.
i. Preferred node
ii. Relocate to preferred member
iii. User script
iv. IP address of the service
v. Device, filesystem & mount information.
vi. Disable on startup.
h. A useful utility for checking the status of the cluster is clustat
