Uninstalling IML
Software Contributor Documentation Table of Contents
Step 1: Uninstall the IML Manager
Stop the manager and httpd
chroma-config stop
Remove the Manager
chkconfig --del chroma-supervisor
yum clean all --enablerepo=*
yum remove chroma* fence-agents*
rm -rf /usr/lib/iml-*
rm -rf /usr/share/chroma-manager/
Optionally, remove nginx
yum remove nginx
Clean up these RPM included files/dirs
rm -rf /usr/lib/python2.7/site-packages/chroma*
Clean up extra yum directories
rm -rf /var/cache/yum/x86_64/7/chroma-manager
Optionally, Remove logs
rm -rf /var/log/chroma
Optionally, export the database
su - postgres
pg_dump -f chromadb.dump chroma
dropdb chroma
logout
Optionally, remove database server
yum remove postgresql*
Optionally stop and remove RabbitMQ
service rabbitmq-server stop
yum remove rabbitmq-server
Note: The IML install creates a repo dir based on a tgz distrubtion and a repo file in /etc/yum.repos.d/ however, both repo dir and repo file are completely removed after installation so they should not need to be deleted now.
Step 2: Uninstall the IML Agent
Remove Host Action and Force Remove Action will do the following on the manager side:
- Remove all manager/agent registrations including sessions and queues to prevent any future communication with the host.
- Remove the hosts cerificate and revoke it’s future use.
- Remove database records for the server and related targets.
The Remove Host Action makes changes to the agent server side.
In the case of ForceRemoveHostJob, the agent side is completely untouched. This is the cause of a few problems and should be used when the communication with the agent is no longer possible. If you do use it, this script can be used to clean up the agent side. You will need to get a shell to the agent machine and issue these commands.
Note also that although the Remove Host Action will remove IML from the agent, it doesn’t do a complete job.
There are some installed artifacts that are not removed. This script can be used to make sure the server is completely uninstalled.
Some of the actions below are optional. If you plan on reinstalling the agent, you may choose not to do the optional items.
Stop and deregister server
service chroma-agent stop
/sbin/chkconfig --del chroma-agent
Remove agent software
yum remove -y chroma-agent chroma-agent-management iml_sos_plugin iml-device-scanner python2-iml-common* lustre-iokit lustre-osd-ldiskfs-mount lustre-osd-zfs-mount
rm -rf /etc/yum.repos.d/Intel-Lustre-Agent.repo
rm -rf /var/lib/chroma/
rm -rf /var/lib/iml/
rm -rf /etc/iml/
rm -rf /etc/yum.repos.d/Intel-Lustre-Agent.repo
rm -rf /usr/lib/python2.7/site-packages/chroma_agent*
Unconfigure NTP
mv -f /etc/ntp.conf.pre-chroma /etc/ntp.conf
Erase all cluster information for this server’s cluster
THIS MEANS THAT OTHER NODES IN THE CLUSTER SHOULD BE REMOVED TOO.
cibadmin -f -E
Kill pacemaker and corosync
systemctl stop pacemaker
systemctl stop corosync
# --OR--
killall -9 pacemaker\; killall -9 corosync # <-- Only if necessary
Reset firewall setting
Get the multicast port from the corosync setting, and used in the iptables command
grep 'mcastport' /etc/corosync/corosync.conf
rm -f /etc/corosync/corosync.conf
Remove firewalld
systemctl status firewalld
systemctl disable firewalld
firewall-cmd --state
-- OR --
/sbin/iptables -D INPUT -m state --state new -p udp --dport MCAST-PORT -j ACCEPT
REMOVE "-A INPUT -m state --state NEW -m udp -p udp --dport MCAST-PORT -j ACCEPT" from /etc/sysconfig/iptables
REMOVE "--port=MCAST-PORT:udp" from /etc/sysconfig/system-config-firewall
remove pacemaker and corosync
yum -y remove pacemaker-* corosync*
rm -f /var/lib/heartbeat/crm/* /var/lib/corosync/*
unconfigure ring1 interface
ifconfig $SERVER_RING1 0.0.0.0 down
rm -f /etc/sysconfig/network-scripts/ifcfg-$SERVER_RING1
unconfigure lnet
rm -f /etc/modprobe.d/iml_lnet_module_parameters.conf
umount targets
umount -a -tlustre -f
Reset your Linux kernel
Check the installed kernel, if the kernel has ‘lustre’ in the name, then uninstall the kernel.
rpm -qR lustre-client-modules | grep 'kernel'
Use Grub to set the desired kernel
awk -F\' '$1=="menuentry " {print i++ " : " $2}' /etc/grub2.cfg
# 0 : CentOS Linux (3.10.0-1062.9.1.el7_lustre.x86_64) 7 (Core)
# 1 : CentOS Linux (3.10.0-514.6.1.el7.x86_64) 7 (Core)
# 2 : CentOS Linux (0-rescue-8018a73b69a84a48bde20d088bca3238) 7 (Core)
grub2-set-default 1
grub2-editenv list
Now that the non-lustre kernel has been selected, reboot the node.
After the system reboots, remove the lustre rpm
rpm -q kernel
Example Output:
kernel-3.10.0-1062.9.1.el7.x86_64
kernel-3.10.0-1062.9.1.el7_lustre.x86_64
This can take a while.
yum remove kernel-3.10.0-1062.9.1.el7_lustre.x86_64