HA Solution Overview

While a xCAT management node xcatmn1 is running as a primary management node, another node - xcatmn2 can be configured to act as primary management node in case xcatmn1 becomes unavailable. The process is manual and requires disabling primary xcatmn1 and activating backup xcatmn2. Both nodes require access to shared storage described below. Use of Virtual IP is also required.

An interactive sample script xcatha.py is available to guide through the steps of disabling and activation of xCAT management nodes. Dryrun option in that scrip allows viewing the actions without executing them.

Configure and Activate Primary xCAT Management Node

Configure Virtual IP

Existing xCAT management node IP should be configured as Virtual IP address, the Virtual IP address should be non-persistent, it needs to be re-configured right after the management node is rebooted. This non-persistent Virtual IP address is designed to avoid ip address conflict when the original primary management node is recovered with this Virtual IP address configured. Since the Virtual IP is non-persistent, the network interface should have a persistent IP address.

  1. Configure another IP on primary management node for network interface as static IP, for example,

    1. Configure as static IP:

      ip addr add dev eth0
    2. Edit ifcfg-eth0 file as:

    3. If want to take new static ip effect immediately, login xcatmn1 using, and restart network service, then add original static IP on primary management node as Virtual IP

      ssh -l root
      service network restart
      ip addr add  brd + dev eth0 label eth0:0
  2. Add into postgresql configuration file on primary management node

    1. Add into /var/lib/pgsql/data/pg_hba.conf:

      host    all          all      md5
    2. Add into listen_addresses variable in /var/lib/pgsql/data/postgresql.conf:

      listen_addresses = 'localhost,,'
  3. Modify provision network entry mgtifname as eth0:0 on primary management node:

    tabedit networks

Configure Shared Data

The following xCAT directory structure should be accessible from primary xCAT management node:


Synchronize /etc/hosts

Since the /etc/hosts is used by xCAT commands, the /etc/hosts should be synchronized between the primary management node and backup management node.

Synchronize Clock

It is recommended that the clocks are synchronized between the primary management node and backup management node.

Activate Primary xCAT Management Node

Use xcatha.py interactive activate xcatmn1:

./xcatha.py -a
[Admin] Verify VIP is configured on this node
Continue? [[Y]es/[N]o]:
[Admin] Verify that the following is configured to be saved in shared storage and accessible from this node:
... /install
... /etc/xcat
... /root/.xcat
... /var/lib/pgsql
... /tftpboot
Continue? [[Y]es/[N]o]:
[xCAT] Starting up services:
... postgresql
... xcatd
... named
... dhcpd
... ntpd
... conserver
... goconserver
Continue? [[Y]es/[N]o/[D]ryrun]:
2018-06-24 22:13:09,428 - INFO - ===> Start all services stage <===
2018-06-24 22:13:10,559 - DEBUG - systemctl start postgresql [Passed]
2018-06-24 22:13:13,298 - DEBUG - systemctl start xcatd [Passed]
2018-06-24 22:13:13,715 - DEBUG - lsdef -t site -i domain|grep domain [Passed]
Handling bybc0607 in /etc/hosts.
Handling localhost in /etc/hosts.
Handling bybc0609 in /etc/hosts.
Handling localhost in /etc/hosts.
Getting reverse zones, this may take several minutes for a large cluster.
Completed getting reverse zones.
Updating zones.
Completed updating zones.
Restarting named
Restarting named complete
Updating DNS records, this may take several minutes for a large cluster.
Completed updating DNS records.
DNS setup is completed
2018-06-24 22:13:17,320 - DEBUG - makedns -n [Passed]
Renamed existing dhcp configuration file to  /etc/dhcp/dhcpd.conf.xcatbak

Warning: No dynamic range specified for If hardware discovery is being used, a dynamic range is required.
2018-06-24 22:13:17,811 - DEBUG - makedhcp -n [Passed]
2018-06-24 22:13:18,746 - DEBUG - makedhcp -a [Passed]
2018-06-24 22:13:18,800 - DEBUG - systemctl start ntpd [Passed]
2018-06-24 22:13:19,353 - DEBUG - makeconservercf [Passed]
2018-06-24 22:13:19,449 - DEBUG - systemctl start conserver [Passed]

Activate Backup xCAT Management Node to be Primary Management Node

  1. Install xCAT on backup xCAT management node xcatmn2 with local disk

  2. Switch to PostgreSQL database

  3. Disable and deactivate services using xcatha.py -d on both xcatmn2 and xcatmn1

  4. Remove Virtual IP from primary xCAT Management Node xcatmn1:

    ip addr del dev eth0:0
  5. Configure Virtual IP on xcatmn2

  6. Add Virtual IP into /etc/hosts file xcatmn1 xcatmn1.cluster.com
  7. Connect the following xCAT directories to shared data on xcatmn2:

  8. Add static management node network interface IP into PostgreSQL configuration file

    1. Add into /var/lib/pgsql/data/pg_hba.conf:

      host    all          all      md5
    2. Add into listen_addresses variable in /var/lib/pgsql/data/postgresql.conf:

      listen_addresses = 'localhost,,,'
  9. Use xcatha.py -a to start all related services on xcatmn2

  10. Modify provision network entry mgtifname as eth0:0:

    tabedit networks

Unplanned failover: primary xCAT management node is not accessible

If primary xCAT management node becomes not accessible before being deactivated and backup xCAT management node is activated, it is recommended that the primary node is disconnected from the network before being rebooted. This will ensure that when services are started on reboot, they do not interfere with the same services running on the backup xCAT management node.