Chapter 3. Pacemaker Tools
3.1. Simplify administration using a cluster shell
In the dark past, configuring Pacemaker required the administrator to read and write XML. In true UNIX style, there were also a number of different commands that specialized in different aspects of querying and updating the cluster.
All of that has been greatly simplified with the creation of unified command-line shells (and GUIs) that hide all the messy XML scaffolding.
These shells take all the individual aspects required for managing and configuring a cluster, and pack them into one simple-to-use command line tool.
They even allow you to queue up several changes at once and commit them all at once.
Two popular command-line shells are pcs
and crmsh
. Clusters from Scratch is based on pcs
because it comes with CentOS, but both have similar functionality. Choosing a shell or GUI is a matter of personal preference and what comes with (and perhaps is supported by) your choice of operating system.
Start by taking some time to familiarize yourself with what pcs
can do.
[root@pcmk-1 ~]# pcs
Usage: pcs [-f file] [-h] [commands]...
Control and configure pacemaker and corosync.
Options:
-h, --help Display usage and exit.
-f file Perform actions on file instead of active CIB.
--debug Print all network traffic and external commands run.
--version Print pcs version information. List pcs capabilities if
--full is specified.
--request-timeout Timeout for each outgoing request to another node in
seconds. Default is 60s.
--force Override checks and errors, the exact behavior depends on
the command. WARNING: Using the --force option is
strongly discouraged unless you know what you are doing.
Commands:
cluster Configure cluster options and nodes.
resource Manage cluster resources.
stonith Manage fence devices.
constraint Manage resource constraints.
property Manage pacemaker properties.
acl Manage pacemaker access control lists.
qdevice Manage quorum device provider on the local host.
quorum Manage cluster quorum settings.
booth Manage booth (cluster ticket manager).
status View cluster status.
config View and manage cluster configuration.
pcsd Manage pcs daemon.
node Manage cluster nodes.
alert Manage pacemaker alerts.
As you can see, the different aspects of cluster management are separated into categories. To discover the functionality available in each of these categories, one can issue the command pcs category
help
. Below is an example of all the options available under the status category.
[root@pcmk-1 ~]# pcs status help
Usage: pcs status [commands]...
View current cluster and resource status
Commands:
[status] [--full | --hide-inactive]
View all information about the cluster and resources (--full provides
more details, --hide-inactive hides inactive resources).
resources [<resource id> | --full | --groups | --hide-inactive]
Show all currently configured resources or if a resource is specified
show the options for the configured resource. If --full is specified,
all configured resource options will be displayed. If --groups is
specified, only show groups (and their resources). If --hide-inactive
is specified, only show active resources.
groups
View currently configured groups and their resources.
cluster
View current cluster status.
corosync
View current membership information as seen by corosync.
quorum
View current quorum status.
qdevice <device model> [--full] [<cluster name>]
Show runtime status of specified model of quorum device provider. Using
--full will give more detailed output. If <cluster name> is specified,
only information about the specified cluster will be displayed.
nodes [corosync | both | config]
View current status of nodes from pacemaker. If 'corosync' is
specified, view current status of nodes from corosync instead. If
'both' is specified, view current status of nodes from both corosync &
pacemaker. If 'config' is specified, print nodes from corosync &
pacemaker configuration.
pcsd [<node>]...
Show current status of pcsd on nodes specified, or on all nodes
configured in the local cluster if no nodes are specified.
xml
View xml version of status (output from crm_mon -r -1 -X).
Additionally, if you are interested in the version and supported cluster stack(s) available with your Pacemaker installation, run:
[root@pcmk-1 ~]# pacemakerd --features
Pacemaker 1.1.18-11.el7_5.3 (Build: 2b07d5c5a9)
Supporting v3.0.14: generated-manpages agent-manpages ncurses libqb-logging libqb-ipc systemd nagios corosync-native atomic-attrd acls
Note
If the SNMP and/or email options are not listed, then Pacemaker was not built to support them. This may be by the choice of your distribution, or the required libraries may not have been available. Please contact whoever supplied you with the packages for more details.
Chapter 4. Start and Verify Cluster
Now that corosync is configured, it is time to start the cluster. The command below will start corosync and pacemaker on both nodes in the cluster. If you are issuing the start command from a different node than the one you ran the pcs cluster auth
command on earlier, you must authenticate on the current node you are logged into before you will be allowed to start the cluster.
[root@pcmk-1 ~]# pcs cluster start --all
pcmk-1: Starting Cluster...
pcmk-2: Starting Cluster...
Note
An alternative to using the pcs cluster start --all
command is to issue either of the below command sequences on each node in the cluster separately:
# pcs cluster start
Starting Cluster...
or
# systemctl start corosync.service
# systemctl start pacemaker.service
Important
In this example, we are not enabling the corosync and pacemaker services to start at boot. If a cluster node fails or is rebooted, you will need to run pcs cluster start nodename
(or --all
) to start the cluster on it. While you could enable the services to start at boot, requiring a manual start of cluster services gives you the opportunity to do a post-mortem investigation of a node failure before returning it to the cluster.
4.2. Verify Corosync Installation
First, use corosync-cfgtool
to check whether cluster communication is happy:
[root@pcmk-1 ~]# corosync-cfgtool -s
Printing ring status.
Local node ID 1
RING ID 0
id = 192.168.122.101
status = ring 0 active with no faults
We can see here that everything appears normal with our fixed IP address (not a 127.0.0.x loopback address) listed as the id, and no faults for the status.
If you see something different, you might want to start by checking the node’s network, firewall and SELinux configurations.
Next, check the membership and quorum APIs:
[root@pcmk-1 ~]# corosync-cmapctl | grep members
runtime.totem.pg.mrp.srp.members.1.config_version (u64) = 0
runtime.totem.pg.mrp.srp.members.1.ip (str) = r(0) ip(192.168.122.101)
runtime.totem.pg.mrp.srp.members.1.join_count (u32) = 1
runtime.totem.pg.mrp.srp.members.1.status (str) = joined
runtime.totem.pg.mrp.srp.members.2.config_version (u64) = 0
runtime.totem.pg.mrp.srp.members.2.ip (str) = r(0) ip(192.168.122.102)
runtime.totem.pg.mrp.srp.members.2.join_count (u32) = 1
runtime.totem.pg.mrp.srp.members.2.status (str) = joined
[root@pcmk-1 ~]# pcs status corosync
Membership information
\----------------------
Nodeid Votes Name
1 1 pcmk-1 (local)
2 1 pcmk-2
You should see both nodes have joined the cluster.
4.3. Verify Pacemaker Installation
Now that we have confirmed that Corosync is functional, we can check the rest of the stack. Pacemaker has already been started, so verify the necessary processes are running:
[root@pcmk-1 ~]# ps axf
PID TTY STAT TIME COMMAND
2 ? S 0:00 [kthreadd]
...lots of processes...
1362 ? Ssl 0:35 corosync
1379 ? Ss 0:00 /usr/sbin/pacemakerd -f
1380 ? Ss 0:00 \_ /usr/libexec/pacemaker/cib
1381 ? Ss 0:00 \_ /usr/libexec/pacemaker/stonithd
1382 ? Ss 0:00 \_ /usr/libexec/pacemaker/lrmd
1383 ? Ss 0:00 \_ /usr/libexec/pacemaker/attrd
1384 ? Ss 0:00 \_ /usr/libexec/pacemaker/pengine
1385 ? Ss 0:00 \_ /usr/libexec/pacemaker/crmd
If that looks OK, check the pcs status
output:
[root@pcmk-1 ~]# pcs status
Cluster name: mycluster
WARNING: no stonith devices and stonith-enabled is not false
Stack: corosync
Current DC: pcmk-2 (version 1.1.18-11.el7_5.3-2b07d5c5a9) - partition with quorum
Last updated: Mon Sep 10 16:37:34 2018
Last change: Mon Sep 10 16:30:53 2018 by hacluster via crmd on pcmk-2
2 nodes configured
0 resources configured
Online: [ pcmk-1 pcmk-2 ]
No resources
Daemon Status:
corosync: active/disabled
pacemaker: active/disabled
pcsd: active/enabled
Finally, ensure there are no start-up errors from corosync or pacemaker (aside from messages relating to not having STONITH configured, which are OK at this point):
[root@pcmk-1 ~]# journalctl -b | grep -i error
Note
Other operating systems may report startup errors in other locations, for example /var/log/messages
.
Repeat these checks on the other node. The results should be the same.
Chapter 5. Create an Active/Passive Cluster
5.1. Explore the Existing Configuration
When Pacemaker starts up, it automatically records the number and details of the nodes in the cluster, as well as which stack is being used and the version of Pacemaker being used.
The first few lines of output should look like this:
[root@pcmk-1 ~]# pcs status
Cluster name: mycluster
WARNING: no stonith devices and stonith-enabled is not false
Stack: corosync
Current DC: pcmk-2 (version 1.1.18-11.el7_5.3-2b07d5c5a9) - partition with quorum
Last updated: Mon Sep 10 16:41:46 2018
Last change: Mon Sep 10 16:30:53 2018 by hacluster via crmd on pcmk-2
2 nodes configured
0 resources configured
Online: [ pcmk-1 pcmk-2 ]
For those who are not of afraid of XML, you can see the raw cluster configuration and status by using the pcs cluster cib
command.
[root@pcmk-1 ~]# pcs cluster cib
<cib crm_feature_set="3.0.14" validate-with="pacemaker-2.10" epoch="5" num_updates="4" admin_epoch="0" cib-last-written="Mon Sep 10 16:30:53 2018" update-origin="pcmk-2" update-client="crmd" update-user="hacluster" have-quorum="1" dc-uuid="2">
<configuration>
<crm_config>
<cluster_property_set id="cib-bootstrap-options">
<nvpair id="cib-bootstrap-options-have-watchdog" name="have-watchdog" value="false"/>
<nvpair id="cib-bootstrap-options-dc-version" name="dc-version" value="1.1.18-11.el7_5.3-2b07d5c5a9"/>
<nvpair id="cib-bootstrap-options-cluster-infrastructure" name="cluster-infrastructure" value="corosync"/>
<nvpair id="cib-bootstrap-options-cluster-name" name="cluster-name" value="mycluster"/>
</cluster_property_set>
</crm_config>
<nodes>
<node id="1" uname="pcmk-1"/>
<node id="2" uname="pcmk-2"/>
</nodes>
<resources/>
<constraints/>
</configuration>
<status>
<node_state id="1" uname="pcmk-1" in_ccm="true" crmd="online" crm-debug-origin="do_state_transition" join="member" expected="member">
<lrm id="1">
<lrm_resources/>
</lrm>
</node_state>
<node_state id="2" uname="pcmk-2" in_ccm="true" crmd="online" crm-debug-origin="do_state_transition" join="member" expected="member">
<lrm id="2">
<lrm_resources/>
</lrm>
</node_state>
</status>
</cib>
Example 5.1. The last XML you’ll see in this document
Before we make any changes, it’s a good idea to check the validity of the configuration.
[root@pcmk-1 ~]# crm_verify -L -V
error: unpack_resources: Resource start-up disabled since no STONITH resources have been defined
error: unpack_resources: Either configure some or disable STONITH with the stonith-enabled option
error: unpack_resources: NOTE: Clusters with shared data need STONITH to ensure data integrity
Errors found during check: config not valid
As you can see, the tool has found some errors.
In order to guarantee the safety of your data, [] fencing (also called STONITH) is enabled by default. However, it also knows when no STONITH configuration has been supplied and reports this as a problem (since the cluster will not be able to make progress if a situation requiring node fencing arises).
We will disable this feature for now and configure it later.
To disable STONITH, set the stonith-enabled cluster option to false:
[root@pcmk-1 ~]# pcs property set stonith-enabled=false
[root@pcmk-1 ~]# crm_verify -L
With the new cluster option set, the configuration is now valid.
Warning
The use of
stonith-enabled=false
is completely inappropriate for a production cluster. It tells the cluster to simply pretend that failed nodes are safely powered off. Some vendors will refuse to support clusters that have STONITH disabled. We disable STONITH here only to defer the discussion of its configuration, which can differ widely from one installation to the next. See
Section 8.1, “What is STONITH?” for information on why STONITH is important and details on how to configure it.
Our first resource will be a unique IP address that the cluster can bring up on either node. Regardless of where any cluster service(s) are running, end users need a consistent address to contact them on. Here, I will choose 192.168.122.120 as the floating address, give it the imaginative name ClusterIP and tell the cluster to check whether it is running every 30 seconds.
Warning
The chosen address must not already be in use on the network. Do not reuse an IP address one of the nodes already has configured.
[root@pcmk-1 ~]# pcs resource create ClusterIP ocf:heartbeat:IPaddr2 \
ip=192.168.122.120 cidr_netmask=32 op monitor interval=30s
Another important piece of information here is ocf:heartbeat:IPaddr2. This tells Pacemaker three things about the resource you want to add:
The first field (ocf in this case) is the standard to which the resource script conforms and where to find it.
The second field (heartbeat in this case) is standard-specific; for OCF resources, it tells the cluster which OCF namespace the resource script is in.
The third field (IPaddr2 in this case) is the name of the resource script.
To obtain a list of the available resource standards (the ocf part of ocf:heartbeat:IPaddr2), run:
[root@pcmk-1 ~]# pcs resource standards
lsb
ocf
service
systemd
To obtain a list of the available OCF resource providers (the heartbeat part of ocf:heartbeat:IPaddr2), run:
[root@pcmk-1 ~]# pcs resource providers
heartbeat
openstack
pacemaker
Finally, if you want to see all the resource agents available for a specific OCF provider (the IPaddr2 part of ocf:heartbeat:IPaddr2), run:
[root@pcmk-1 ~]# pcs resource agents ocf:heartbeat
apache
aws-vpc-move-ip
awseip
awsvip
azure-lb
clvm
.
. (skipping lots of resources to save space)
.
symlink
tomcat
VirtualDomain
Xinetd
Now, verify that the IP resource has been added, and display the cluster’s status to see that it is now active:
[root@pcmk-1 ~]# pcs status
Cluster name: mycluster
Stack: corosync
Current DC: pcmk-2 (version 1.1.18-11.el7_5.3-2b07d5c5a9) - partition with quorum
Last updated: Mon Sep 10 16:55:26 2018
Last change: Mon Sep 10 16:53:42 2018 by root via cibadmin on pcmk-1
2 nodes configured
1 resource configured
Online: [ pcmk-1 pcmk-2 ]
Full list of resources:
ClusterIP (ocf::heartbeat:IPaddr2): Started pcmk-1
Daemon Status:
corosync: active/disabled
pacemaker: active/disabled
pcsd: active/enabled
Since our ultimate goal is high availability, we should test failover of our new resource before moving on.
First, find the node on which the IP address is running.
[root@pcmk-1 ~]# pcs status
Cluster name: mycluster
Stack: corosync
Current DC: pcmk-2 (version 1.1.18-11.el7_5.3-2b07d5c5a9) - partition with quorum
Last updated: Mon Sep 10 16:55:26 2018
Last change: Mon Sep 10 16:53:42 2018 by root via cibadmin on pcmk-1
2 nodes configured
1 resource configured
Online: [ pcmk-1 pcmk-2 ]
Full list of resources:
ClusterIP (ocf::heartbeat:IPaddr2): Started pcmk-1
You can see that the status of the ClusterIP resource is Started on a particular node (in this example, pcmk-1). Shut down Pacemaker and Corosync on that machine to trigger a failover.
[root@pcmk-1 ~]# pcs cluster stop pcmk-1
Stopping Cluster (pacemaker)...
Stopping Cluster (corosync)...
Note
A cluster command such as pcs cluster stop nodename
can be run from any node in the cluster, not just the affected node.
Verify that pacemaker and corosync are no longer running:
[root@pcmk-1 ~]# pcs status
Error: cluster is not currently running on this node
Go to the other node, and check the cluster status.
[root@pcmk-2 ~]# pcs status
Cluster name: mycluster
Stack: corosync
Current DC: pcmk-2 (version 1.1.18-11.el7_5.3-2b07d5c5a9) - partition with quorum
Last updated: Mon Sep 10 16:57:22 2018
Last change: Mon Sep 10 16:53:42 2018 by root via cibadmin on pcmk-1
2 nodes configured
1 resource configured
Online: [ pcmk-2 ]
OFFLINE: [ pcmk-1 ]
Full list of resources:
ClusterIP (ocf::heartbeat:IPaddr2): Started pcmk-2
Daemon Status:
corosync: active/disabled
pacemaker: active/disabled
pcsd: active/enabled
Notice that pcmk-1 is OFFLINE for cluster purposes (its pcsd is still active, allowing it to receive pcs
commands, but it is not participating in the cluster).
Also notice that ClusterIP is now running on pcmk-2 — failover happened automatically, and no errors are reported.
Quorum
If a cluster splits into two (or more) groups of nodes that can no longer communicate with each other (aka. partitions), quorum is used to prevent resources from starting on more nodes than desired, which would risk data corruption.
A cluster has quorum when more than half of all known nodes are online in the same partition, or for the mathematically inclined, whenever the following equation is true:
total_nodes < 2 * active_nodes
For example, if a 5-node cluster split into 3- and 2-node paritions, the 3-node partition would have quorum and could continue serving resources. If a 6-node cluster split into two 3-node partitions, neither partition would have quorum; pacemaker’s default behavior in such cases is to stop all resources, in order to prevent data corruption.
Two-node clusters are a special case. By the above definition, a two-node cluster would only have quorum when both nodes are running. This would make the creation of a two-node cluster pointless, but corosync has the ability to treat two-node clusters as if only one node is required for quorum.
The pcs cluster setup
command will automatically configure two_node: 1 in corosync.conf
, so a two-node cluster will "just work".
If you are using a different cluster shell, you will have to configure corosync.conf
appropriately yourself.
Now, simulate node recovery by restarting the cluster stack on pcmk-1, and check the cluster’s status. (It may take a little while before the cluster gets going on the node, but it eventually will look like the below.)
[root@pcmk-1 ~]# pcs cluster start pcmk-1
pcmk-1: Starting Cluster...
[root@pcmk-1 ~]# pcs status
Cluster name: mycluster
Stack: corosync
Current DC: pcmk-2 (version 1.1.18-11.el7_5.3-2b07d5c5a9) - partition with quorum
Last updated: Mon Sep 10 17:00:04 2018
Last change: Mon Sep 10 16:53:42 2018 by root via cibadmin on pcmk-1
2 nodes configured
1 resource configured
Online: [ pcmk-1 pcmk-2 ]
Full list of resources:
ClusterIP (ocf::heartbeat:IPaddr2): Started pcmk-2
Daemon Status:
corosync: active/disabled
pacemaker: active/disabled
pcsd: active/enabled
5.4. Prevent Resources from Moving after Recovery
In most circumstances, it is highly desirable to prevent healthy resources from being moved around the cluster. Moving resources almost always requires a period of downtime. For complex services such as databases, this period can be quite long.
To address this, Pacemaker has the concept of resource stickiness, which controls how strongly a service prefers to stay running where it is. You may like to think of it as the "cost" of any downtime. By default, Pacemaker assumes there is zero cost associated with moving resources and will do so to achieve "optimal" [] resource placement. We can specify a different stickiness for every resource, but it is often sufficient to change the default.
[root@pcmk-1 ~]# pcs resource defaults resource-stickiness=100
Warning: Defaults do not apply to resources which override them with their own defined values
[root@pcmk-1 ~]# pcs resource defaults
resource-stickiness: 100
Chapter 6. Add Apache HTTP Server as a Cluster Service
Now that we have a basic but functional active/passive two-node cluster, we’re ready to add some real services. We’re going to start with Apache HTTP Server because it is a feature of many clusters and relatively simple to configure.
Before continuing, we need to make sure Apache is installed on both hosts. We also need the wget tool in order for the cluster to be able to check the status of the Apache server.
# yum install -y httpd wget
# firewall-cmd --permanent --add-service=http
# firewall-cmd --reload
Important
Do not enable the httpd service. Services that are intended to be managed via the cluster software should never be managed by the OS. It is often useful, however, to manually start the service, verify that it works, then stop it again, before adding it to the cluster. This allows you to resolve any non-cluster-related problems before continuing. Since this is a simple example, we’ll skip that step here.
6.2. Create Website Documents
We need to create a page for Apache to serve. On CentOS 7.5, the default Apache document root is /var/www/html, so we’ll create an index file there. For the moment, we will simplify things by serving a static site and manually synchronizing the data between the two nodes, so run this command on both nodes:
# cat <<-END >/var/www/html/index.html
<html>
<body>My Test Site - $(hostname)</body>
</html>
END
6.3. Enable the Apache status URL
In order to monitor the health of your Apache instance, and recover it if it fails, the resource agent used by Pacemaker assumes the server-status URL is available. On both nodes, enable the URL with:
# cat <<-END >/etc/httpd/conf.d/status.conf
<Location /server-status>
SetHandler server-status
Require local
</Location>
END
Note
If you are using a different operating system, server-status may already be enabled or may be configurable in a different location. If you are using a version of Apache HTTP Server less than 2.4, the syntax will be different.
At this point, Apache is ready to go, and all that needs to be done is to add it to the cluster. Let’s call the resource WebSite. We need to use an OCF resource script called apache in the heartbeat namespace. [] The script’s only required parameter is the path to the main Apache configuration file, and we’ll tell the cluster to check once a minute that Apache is still running.
[root@pcmk-1 ~]# pcs resource create WebSite ocf:heartbeat:apache \
configfile=/etc/httpd/conf/httpd.conf \
statusurl="http://localhost/server-status" \
op monitor interval=1min
By default, the operation timeout for all resources' start, stop, and monitor operations is 20 seconds. In many cases, this timeout period is less than a particular resource’s advised timeout period. For the purposes of this tutorial, we will adjust the global operation timeout default to 240 seconds.
[root@pcmk-1 ~]# pcs resource op defaults timeout=240s
Warning: Defaults do not apply to resources which override them with their own defined values
[root@pcmk-1 ~]# pcs resource op defaults
timeout: 240s
Note
In a production cluster, it is usually better to adjust each resource’s start, stop, and monitor timeouts to values that are appropriate to the behavior observed in your environment, rather than adjust the global default.
After a short delay, we should see the cluster start Apache.
[root@pcmk-1 ~]# pcs status
Cluster name: mycluster
Stack: corosync
Current DC: pcmk-2 (version 1.1.18-11.el7_5.3-2b07d5c5a9) - partition with quorum
Last updated: Mon Sep 10 17:06:22 2018
Last change: Mon Sep 10 17:05:41 2018 by root via cibadmin on pcmk-1
2 nodes configured
2 resources configured
Online: [ pcmk-1 pcmk-2 ]
Full list of resources:
ClusterIP (ocf::heartbeat:IPaddr2): Started pcmk-2
WebSite (ocf::heartbeat:apache): Started pcmk-1
Daemon Status:
corosync: active/disabled
pacemaker: active/disabled
pcsd: active/enabled
Wait a moment, the WebSite resource isn’t running on the same host as our IP address!
Note
If, in the pcs status
output, you see the WebSite resource has failed to start, then you’ve likely not enabled the status URL correctly. You can check whether this is the problem by running:
wget -O - http://localhost/server-status
If you see Not Found or Forbidden in the output, then this is likely the problem. Ensure that the <Location /server-status> block is correct.
6.5. Ensure Resources Run on the Same Host
To reduce the load on any one machine, Pacemaker will generally try to spread the configured resources across the cluster nodes. However, we can tell the cluster that two resources are related and need to run on the same host (or not at all). Here, we instruct the cluster that WebSite can only run on the host that ClusterIP is active on.
To achieve this, we use a colocation constraint that indicates it is mandatory for WebSite to run on the same node as ClusterIP. The "mandatory" part of the colocation constraint is indicated by using a score of INFINITY. The INFINITY score also means that if ClusterIP is not active anywhere, WebSite will not be permitted to run.
Note
If ClusterIP is not active anywhere, WebSite will not be permitted to run anywhere.
Important
Colocation constraints are "directional", in that they imply certain things about the order in which the two resources will have a location chosen. In this case, we’re saying that WebSite needs to be placed on the same machine as ClusterIP, which implies that the cluster must know the location of ClusterIP before choosing a location for WebSite.
[root@pcmk-1 ~]# pcs constraint colocation add WebSite with ClusterIP INFINITY
[root@pcmk-1 ~]# pcs constraint
Location Constraints:
Ordering Constraints:
Colocation Constraints:
WebSite with ClusterIP (score:INFINITY)
Ticket Constraints:
[root@pcmk-1 ~]# pcs status
Cluster name: mycluster
Stack: corosync
Current DC: pcmk-2 (version 1.1.18-11.el7_5.3-2b07d5c5a9) - partition with quorum
Last updated: Mon Sep 10 17:08:54 2018
Last change: Mon Sep 10 17:08:27 2018 by root via cibadmin on pcmk-1
2 nodes configured
2 resources configured
Online: [ pcmk-1 pcmk-2 ]
Full list of resources:
ClusterIP (ocf::heartbeat:IPaddr2): Started pcmk-2
WebSite (ocf::heartbeat:apache): Started pcmk-2
Daemon Status:
corosync: active/disabled
pacemaker: active/disabled
pcsd: active/enabled
6.6. Ensure Resources Start and Stop in Order
Like many services, Apache can be configured to bind to specific IP addresses on a host or to the wildcard IP address. If Apache binds to the wildcard, it doesn’t matter whether an IP address is added before or after Apache starts; Apache will respond on that IP just the same. However, if Apache binds only to certain IP address(es), the order matters: If the address is added after Apache starts, Apache won’t respond on that address.
To be sure our WebSite responds regardless of Apache’s address configuration, we need to make sure ClusterIP not only runs on the same node, but starts before WebSite. A colocation constraint only ensures the resources run together, not the order in which they are started and stopped.
We do this by adding an ordering constraint. By default, all order constraints are mandatory, which means that the recovery of ClusterIP will also trigger the recovery of WebSite.
[root@pcmk-1 ~]# pcs constraint order ClusterIP then WebSite
Adding ClusterIP WebSite (kind: Mandatory) (Options: first-action=start then-action=start)
[root@pcmk-1 ~]# pcs constraint
Location Constraints:
Ordering Constraints:
start ClusterIP then start WebSite (kind:Mandatory)
Colocation Constraints:
WebSite with ClusterIP (score:INFINITY)
Ticket Constraints:
6.7. Prefer One Node Over Another
Pacemaker does not rely on any sort of hardware symmetry between nodes, so it may well be that one machine is more powerful than the other.
In such cases, you may want to host the resources on the more powerful node when it is available, to have the best performance — or you may want to host the resources on the less powerful node when it’s available, so you don’t have to worry about whether you can handle the load after a failover.
To do this, we create a location constraint.
In the location constraint below, we are saying the WebSite resource prefers the node pcmk-1 with a score of 50. Here, the score indicates how strongly we’d like the resource to run at this location.
[root@pcmk-1 ~]# pcs constraint location WebSite prefers pcmk-1=50
[root@pcmk-1 ~]# pcs constraint
Location Constraints:
Resource: WebSite
Enabled on: pcmk-1 (score:50)
Ordering Constraints:
start ClusterIP then start WebSite (kind:Mandatory)
Colocation Constraints:
WebSite with ClusterIP (score:INFINITY)
Ticket Constraints:
[root@pcmk-1 ~]# pcs status
Cluster name: mycluster
Stack: corosync
Current DC: pcmk-2 (version 1.1.18-11.el7_5.3-2b07d5c5a9) - partition with quorum
Last updated: Mon Sep 10 17:21:41 2018
Last change: Mon Sep 10 17:21:14 2018 by root via cibadmin on pcmk-1
2 nodes configured
2 resources configured
Online: [ pcmk-1 pcmk-2 ]
Full list of resources:
ClusterIP (ocf::heartbeat:IPaddr2): Started pcmk-2
WebSite (ocf::heartbeat:apache): Started pcmk-2
Daemon Status:
corosync: active/disabled
pacemaker: active/disabled
pcsd: active/enabled
Wait a minute, the resources are still on pcmk-2!
Even though WebSite now prefers to run on pcmk-1, that preference is (intentionally) less than the resource stickiness (how much we preferred not to have unnecessary downtime).
To see the current placement scores, you can use a tool called crm_simulate.
[root@pcmk-1 ~]# crm_simulate -sL
Current cluster status:
Online: [ pcmk-1 pcmk-2 ]
ClusterIP (ocf::heartbeat:IPaddr2): Started pcmk-2
WebSite (ocf::heartbeat:apache): Started pcmk-2
Allocation scores:
native_color: ClusterIP allocation score on pcmk-1: 50
native_color: ClusterIP allocation score on pcmk-2: 200
native_color: WebSite allocation score on pcmk-1: -INFINITY
native_color: WebSite allocation score on pcmk-2: 100
Transition Summary:
6.8. Move Resources Manually
There are always times when an administrator needs to override the cluster and force resources to move to a specific location. In this example, we will force the WebSite to move to pcmk-1.
We will use the pcs resource move command to create a temporary constraint with a score of INFINITY. While we could update our existing constraint, using move allows to easily get rid of the temporary constraint later. If desired, we could even give a lifetime for the constraint, so it would expire automatically — but we don’t that in this example.
[root@pcmk-1 ~]# pcs resource move WebSite pcmk-1
[root@pcmk-1 ~]# pcs constraint
Location Constraints:
Resource: WebSite
Enabled on: pcmk-1 (score:50)
Enabled on: pcmk-1 (score:INFINITY) (role: Started)
Ordering Constraints:
start ClusterIP then start WebSite (kind:Mandatory)
Colocation Constraints:
WebSite with ClusterIP (score:INFINITY)
Ticket Constraints:
[root@pcmk-1 ~]# pcs status
Cluster name: mycluster
Stack: corosync
Current DC: pcmk-2 (version 1.1.18-11.el7_5.3-2b07d5c5a9) - partition with quorum
Last updated: Mon Sep 10 17:28:55 2018
Last change: Mon Sep 10 17:28:27 2018 by root via crm_resource on pcmk-1
2 nodes configured
2 resources configured
Online: [ pcmk-1 pcmk-2 ]
Full list of resources:
ClusterIP (ocf::heartbeat:IPaddr2): Started pcmk-1
WebSite (ocf::heartbeat:apache): Started pcmk-1
Daemon Status:
corosync: active/disabled
pacemaker: active/disabled
pcsd: active/enabled
Once we’ve finished whatever activity required us to move the resources to pcmk-1 (in our case nothing), we can then allow the cluster to resume normal operation by removing the new constraint. Due to our first location constraint and our default stickiness, the resources will remain on pcmk-1.
We will use the pcs resource clear command, which removes all temporary constraints previously created by pcs resource move or pcs resource ban.
[root@pcmk-1 ~]# pcs resource clear WebSite
[root@pcmk-1 ~]# pcs constraint
Location Constraints:
Resource: WebSite
Enabled on: pcmk-1 (score:50)
Ordering Constraints:
start ClusterIP then start WebSite (kind:Mandatory)
Colocation Constraints:
WebSite with ClusterIP (score:INFINITY)
Ticket Constraints:
Note that the INFINITY location constraint is now gone. If we check the cluster status, we can also see that (as expected) the resources are still active on pcmk-1.
[root@pcmk-1 ~]# pcs status
Cluster name: mycluster
Stack: corosync
Current DC: pcmk-2 (version 1.1.18-11.el7_5.3-2b07d5c5a9) - partition with quorum
Last updated: Mon Sep 10 17:31:47 2018
Last change: Mon Sep 10 17:31:04 2018 by root via crm_resource on pcmk-1
2 nodes configured
2 resources configured
Online: [ pcmk-1 pcmk-2 ]
Full list of resources:
ClusterIP (ocf::heartbeat:IPaddr2): Started pcmk-1
WebSite (ocf::heartbeat:apache): Started pcmk-1
Daemon Status:
corosync: active/disabled
pacemaker: active/disabled
pcsd: active/enabled
To remove the constraint with the score of 50, we would first get the constraint’s ID using pcs constraint --full, then remove it with pcs constraint remove and the ID. We won’t show those steps here, but feel free to try it on your own, with the help of the pcs man page if necessary.
Chapter 7. Replicate Storage Using DRBD
Even if you’re serving up static websites, having to manually synchronize the contents of that website to all the machines in the cluster is not ideal. For dynamic websites, such as a wiki, it’s not even an option. Not everyone care afford network-attached storage, but somehow the data needs to be kept in sync.
Enter DRBD, which can be thought of as network-based RAID-1. []
7.1. Install the DRBD Packages
DRBD itself is included in the upstream kernel,[] but we do need some utilities to use it effectively.
CentOS does not ship these utilities, so we need to enable a third-party repository to get them. Supported packages for many OSes are available from DRBD’s maker
LINBIT, but here we’ll use the free
ELRepo repository.
On both nodes, import the ELRepo package signing key, and enable the repository:
# rpm --import https://www.elrepo.org/RPM-GPG-KEY-elrepo.org
# rpm -Uvh http://www.elrepo.org/elrepo-release-7.0-3.el7.elrepo.noarch.rpm
Retrieving http://www.elrepo.org/elrepo-release-7.0-3.el7.elrepo.noarch.rpm
Preparing... ################################# [100%]
Updating / installing...
1:elrepo-release-7.0-3.el7.elrepo ################################# [100%]
Now, we can install the DRBD kernel module and utilities:
# yum install -y kmod-drbd84 drbd84-utils
DRBD will not be able to run under the default SELinux security policies. If you are familiar with SELinux, you can modify the policies in a more fine-grained manner, but here we will simply exempt DRBD processes from SELinux control:
# semanage permissive -a drbd_t
We will configure DRBD to use port 7789, so allow that port from each host to the other:
[root@pcmk-1 ~]# firewall-cmd --permanent --add-rich-rule='rule family="ipv4" \
source address="192.168.122.102" port port="7789" protocol="tcp" accept'
success
[root@pcmk-1 ~]# firewall-cmd --reload
success
[root@pcmk-2 ~]# firewall-cmd --permanent --add-rich-rule='rule family="ipv4" \
source address="192.168.122.101" port port="7789" protocol="tcp" accept'
success
[root@pcmk-2 ~]# firewall-cmd --reload
success
Note
In this example, we have only two nodes, and all network traffic is on the same LAN. In production, it is recommended to use a dedicated, isolated network for cluster-related traffic, so the firewall configuration would likely be different; one approach would be to add the dedicated network interfaces to the trusted zone.
7.2. Allocate a Disk Volume for DRBD
DRBD will need its own block device on each node. This can be a physical disk partition or logical volume, of whatever size you need for your data. For this document, we will use a 512MiB logical volume, which is more than sufficient for a single HTML file and (later) GFS2 metadata.
[root@pcmk-1 ~]# vgdisplay | grep -e Name -e Free
VG Name centos_pcmk-1
Free PE / Size 255 / 1020.00 MiB
[root@pcmk-1 ~]# lvcreate --name drbd-demo --size 512M centos_pcmk-1
Logical volume "drbd-demo" created.
[root@pcmk-1 ~]# lvs
LV VG Attr LSize Pool Origin Data% Meta% Move Log Cpy%Sync Convert
drbd-demo centos_pcmk-1 -wi-a----- 512.00m
root centos_pcmk-1 -wi-ao---- 3.00g
swap centos_pcmk-1 -wi-ao---- 1.00g
Repeat for the second node, making sure to use the same size:
[root@pcmk-1 ~]# ssh pcmk-2 -- lvcreate --name drbd-demo --size 512M centos_pcmk-2
Logical volume "drbd-demo" created.
With the configuration in place, we can now get DRBD running.
These commands create the local metadata for the DRBD resource, ensure the DRBD kernel module is loaded, and bring up the DRBD resource. Run them on one node:
[root@pcmk-1 ~]# drbdadm create-md wwwdata
--== Thank you for participating in the global usage survey ==--
The server's response is:
you are the 2147th user to install this version
initializing activity log
initializing bitmap (16 KB) to all zero
Writing meta data...
New drbd meta data block successfully created.
success
[root@pcmk-1 ~]# modprobe drbd
[root@pcmk-1 ~]# drbdadm up wwwdata
--== Thank you for participating in the global usage survey ==--
The server's response is:
We can confirm DRBD’s status on this node:
[root@pcmk-1 ~]# cat /proc/drbd
version: 8.4.11-1 (api:1/proto:86-101)
GIT-hash: 66145a308421e9c124ec391a7848ac20203bb03c build by mockbuild@, 2018-04-26 12:10:42
1: cs:WFConnection ro:Secondary/Unknown ds:Inconsistent/DUnknown C r----s
ns:0 nr:0 dw:0 dr:0 al:8 bm:0 lo:0 pe:0 ua:0 ap:0 ep:1 wo:f oos:524236
Because we have not yet initialized the data, this node’s data is marked as Inconsistent. Because we have not yet initialized the second node, the local state is WFConnection (waiting for connection), and the partner node’s status is marked as Unknown.
Now, repeat the above commands on the second node, starting with creating wwwdata.res. After giving it time to connect, when we check the status, it shows:
[root@pcmk-2 ~]# cat /proc/drbd
version: 8.4.11-1 (api:1/proto:86-101)
GIT-hash: 66145a308421e9c124ec391a7848ac20203bb03c build by mockbuild@, 2018-04-26 12:10:42
1: cs:Connected ro:Secondary/Secondary ds:Inconsistent/Inconsistent C r-----
ns:0 nr:0 dw:0 dr:0 al:8 bm:0 lo:0 pe:0 ua:0 ap:0 ep:1 wo:f oos:524236
You can see the state has changed to Connected, meaning the two DRBD nodes are communicating properly, and both nodes are in Secondary role with Inconsistent data.
To make the data consistent, we need to tell DRBD which node should be considered to have the correct data. In this case, since we are creating a new resource, both have garbage, so we’ll just pick pcmk-1 and run this command on it:
[root@pcmk-1 ~]# drbdadm primary --force wwwdata
Note
If you are using a different version of DRBD, the required syntax may be different. See the documentation for your version for how to perform these commands.
If we check the status immediately, we’ll see something like this:
[root@pcmk-1 ~]# cat /proc/drbd
version: 8.4.11-1 (api:1/proto:86-101)
GIT-hash: 66145a308421e9c124ec391a7848ac20203bb03c build by mockbuild@, 2018-04-26 12:10:42
1: cs:SyncSource ro:Primary/Secondary ds:UpToDate/Inconsistent C r-----
ns:43184 nr:0 dw:0 dr:45312 al:8 bm:0 lo:0 pe:0 ua:0 ap:0 ep:1 wo:f oos:481052
[>...................] sync'ed: 8.6% (481052/524236)K
finish: 0:01:51 speed: 4,316 (4,316) K/sec
We can see that this node has the Primary role, the partner node has the Secondary role, this node’s data is now considered UpToDate, the partner node’s data is still Inconsistent, and a progress bar shows how far along the partner node is in synchronizing the data.
After a while, the sync should finish, and you’ll see something like:
[root@pcmk-1 ~]# cat /proc/drbd
version: 8.4.11-1 (api:1/proto:86-101)
GIT-hash: 66145a308421e9c124ec391a7848ac20203bb03c build by mockbuild@, 2018-04-26 12:10:42
1: cs:Connected ro:Primary/Secondary ds:UpToDate/UpToDate C r-----
ns:524236 nr:0 dw:0 dr:526364 al:8 bm:0 lo:0 pe:0 ua:0 ap:0 ep:1 wo:f oos:0
Both sets of data are now UpToDate, and we can proceed to creating and populating a filesystem for our WebSite resource’s documents.
7.5. Populate the DRBD Disk
On the node with the primary role (pcmk-1 in this example), create a filesystem on the DRBD device:
[root@pcmk-1 ~]# mkfs.xfs /dev/drbd1
meta-data=/dev/drbd1 isize=512 agcount=4, agsize=32765 blks
= sectsz=512 attr=2, projid32bit=1
= crc=1 finobt=0, sparse=0
data = bsize=4096 blocks=131059, imaxpct=25
= sunit=0 swidth=0 blks
naming =version 2 bsize=4096 ascii-ci=0 ftype=1
log =internal log bsize=4096 blocks=855, version=2
= sectsz=512 sunit=0 blks, lazy-count=1
realtime =none extsz=4096 blocks=0, rtextents=0
Note
In this example, we create an xfs filesystem with no special options. In a production environment, you should choose a filesystem type and options that are suitable for your application.
Mount the newly created filesystem, populate it with our web document, give it the same SELinux policy as the web document root, then unmount it (the cluster will handle mounting and unmounting it later):
[root@pcmk-1 ~]# mount /dev/drbd1 /mnt
[root@pcmk-1 ~]# cat <<-END >/mnt/index.html
<html>
<body>My Test Site - DRBD</body>
</html>
END
[root@pcmk-1 ~]# chcon -R --reference=/var/www/html /mnt
[root@pcmk-1 ~]# umount /dev/drbd1
One handy feature pcs
has is the ability to queue up several changes into a file and commit those changes all at once. To do this, start by populating the file with the current raw XML config from the CIB.
[root@pcmk-1 ~]# pcs cluster cib drbd_cfg
Using pcs’s -f
option, make changes to the configuration saved in the drbd_cfg
file. These changes will not be seen by the cluster until the drbd_cfg
file is pushed into the live cluster’s CIB later.
Here, we create a cluster resource for the DRBD device, and an additional clone resource to allow the resource to run on both nodes at the same time.
[root@pcmk-1 ~]# pcs -f drbd_cfg resource create WebData ocf:linbit:drbd \
drbd_resource=wwwdata op monitor interval=60s
[root@pcmk-1 ~]# pcs -f drbd_cfg resource master WebDataClone WebData \
master-max=1 master-node-max=1 clone-max=2 clone-node-max=1 \
notify=true
[root@pcmk-1 ~]# pcs -f drbd_cfg resource show
ClusterIP (ocf::heartbeat:IPaddr2): Started pcmk-1
WebSite (ocf::heartbeat:apache): Started pcmk-1
Master/Slave Set: WebDataClone [WebData]
Stopped: [ pcmk-1 pcmk-2 ]
After you are satisfied with all the changes, you can commit them all at once by pushing the drbd_cfg file into the live CIB.
[root@pcmk-1 ~]# pcs cluster cib-push drbd_cfg --config
CIB updated
Let’s see what the cluster did with the new configuration:
[root@pcmk-1 ~]# pcs status
Cluster name: mycluster
Stack: corosync
Current DC: pcmk-2 (version 1.1.18-11.el7_5.3-2b07d5c5a9) - partition with quorum
Last updated: Mon Sep 10 17:58:07 2018
Last change: Mon Sep 10 17:57:53 2018 by root via cibadmin on pcmk-1
2 nodes configured
4 resources configured
Online: [ pcmk-1 pcmk-2 ]
Full list of resources:
ClusterIP (ocf::heartbeat:IPaddr2): Started pcmk-1
WebSite (ocf::heartbeat:apache): Started pcmk-1
Master/Slave Set: WebDataClone [WebData]
Masters: [ pcmk-1 ]
Slaves: [ pcmk-2 ]
Daemon Status:
corosync: active/disabled
pacemaker: active/disabled
pcsd: active/enabled
We can see that WebDataClone (our DRBD device) is running as master (DRBD’s primary role) on pcmk-1 and slave (DRBD’s secondary role) on pcmk-2.
Important
The resource agent should load the DRBD module when needed if it’s not already loaded. If that does not happen, configure your operating system to load the module at boot time. For CentOS 7.5, you would run this on both nodes:
# echo drbd >/etc/modules-load.d/drbd.conf
Now that we have a working DRBD device, we need to mount its filesystem.
In addition to defining the filesystem, we also need to tell the cluster where it can be located (only on the DRBD Primary) and when it is allowed to start (after the Primary was promoted).
We are going to take a shortcut when creating the resource this time. Instead of explicitly saying we want the ocf:heartbeat:Filesystem script, we are only going to ask for Filesystem. We can do this because we know there is only one resource script named Filesystem available to pacemaker, and that pcs is smart enough to fill in the ocf:heartbeat: portion for us correctly in the configuration. If there were multiple Filesystem scripts from different OCF providers, we would need to specify the exact one we wanted.
Once again, we will queue our changes to a file and then push the new configuration to the cluster as the final step.
[root@pcmk-1 ~]# pcs cluster cib fs_cfg
[root@pcmk-1 ~]# pcs -f fs_cfg resource create WebFS Filesystem \
device="/dev/drbd1" directory="/var/www/html" fstype="xfs"
Assumed agent name 'ocf:heartbeat:Filesystem' (deduced from 'Filesystem')
[root@pcmk-1 ~]# pcs -f fs_cfg constraint colocation add \
WebFS with WebDataClone INFINITY with-rsc-role=Master
[root@pcmk-1 ~]# pcs -f fs_cfg constraint order \
promote WebDataClone then start WebFS
Adding WebDataClone WebFS (kind: Mandatory) (Options: first-action=promote then-action=start)
We also need to tell the cluster that Apache needs to run on the same machine as the filesystem and that it must be active before Apache can start.
[root@pcmk-1 ~]# pcs -f fs_cfg constraint colocation add WebSite with WebFS INFINITY
[root@pcmk-1 ~]# pcs -f fs_cfg constraint order WebFS then WebSite
Adding WebFS WebSite (kind: Mandatory) (Options: first-action=start then-action=start)
Review the updated configuration.
[root@pcmk-1 ~]# pcs -f fs_cfg constraint
Location Constraints:
Resource: WebSite
Enabled on: pcmk-1 (score:50)
Ordering Constraints:
start ClusterIP then start WebSite (kind:Mandatory)
promote WebDataClone then start WebFS (kind:Mandatory)
start WebFS then start WebSite (kind:Mandatory)
Colocation Constraints:
WebSite with ClusterIP (score:INFINITY)
WebFS with WebDataClone (score:INFINITY) (with-rsc-role:Master)
WebSite with WebFS (score:INFINITY)
Ticket Constraints:
[root@pcmk-1 ~]# pcs -f fs_cfg resource show
ClusterIP (ocf::heartbeat:IPaddr2): Started pcmk-1
WebSite (ocf::heartbeat:apache): Started pcmk-1
Master/Slave Set: WebDataClone [WebData]
Masters: [ pcmk-1 ]
Slaves: [ pcmk-2 ]
WebFS (ocf::heartbeat:Filesystem): Stopped
After reviewing the new configuration, upload it and watch the cluster put it into effect.
[root@pcmk-1 ~]# pcs cluster cib-push fs_cfg --config
CIB updated
[root@pcmk-1 ~]# pcs status
Cluster name: mycluster
Stack: corosync
Current DC: pcmk-2 (version 1.1.18-11.el7_5.3-2b07d5c5a9) - partition with quorum
Last updated: Mon Sep 10 18:02:24 2018
Last change: Mon Sep 10 18:02:14 2018 by root via cibadmin on pcmk-1
2 nodes configured
5 resources configured
Online: [ pcmk-1 pcmk-2 ]
Full list of resources:
ClusterIP (ocf::heartbeat:IPaddr2): Started pcmk-1
WebSite (ocf::heartbeat:apache): Started pcmk-1
Master/Slave Set: WebDataClone [WebData]
Masters: [ pcmk-1 ]
Slaves: [ pcmk-2 ]
WebFS (ocf::heartbeat:Filesystem): Started pcmk-1
Daemon Status:
corosync: active/disabled
pacemaker: active/disabled
pcsd: active/enabled
7.8. Test Cluster Failover
Previously, we used pcs cluster stop pcmk-1
to stop all cluster services on pcmk-1, failing over the cluster resources, but there is another way to safely simulate node failure.
We can put the node into standby mode. Nodes in this state continue to run corosync and pacemaker but are not allowed to run resources. Any resources found active there will be moved elsewhere. This feature can be particularly useful when performing system administration tasks such as updating packages used by cluster resources.
Put the active node into standby mode, and observe the cluster move all the resources to the other node. The node’s status will change to indicate that it can no longer host resources, and eventually all the resources will move.
[root@pcmk-1 ~]# pcs cluster standby pcmk-1
[root@pcmk-1 ~]# pcs status
Cluster name: mycluster
Stack: corosync
Current DC: pcmk-2 (version 1.1.18-11.el7_5.3-2b07d5c5a9) - partition with quorum
Last updated: Mon Sep 10 18:04:22 2018
Last change: Mon Sep 10 18:03:43 2018 by root via cibadmin on pcmk-1
2 nodes configured
5 resources configured
Node pcmk-1: standby
Online: [ pcmk-2 ]
Full list of resources:
ClusterIP (ocf::heartbeat:IPaddr2): Started pcmk-2
WebSite (ocf::heartbeat:apache): Started pcmk-2
Master/Slave Set: WebDataClone [WebData]
Masters: [ pcmk-2 ]
Stopped: [ pcmk-1 ]
WebFS (ocf::heartbeat:Filesystem): Started pcmk-2
Daemon Status:
corosync: active/disabled
pacemaker: active/disabled
pcsd: active/enabled
Once we’ve done everything we needed to on pcmk-1 (in this case nothing, we just wanted to see the resources move), we can allow the node to be a full cluster member again.
[root@pcmk-1 ~]# pcs cluster unstandby pcmk-1
[root@pcmk-1 ~]# pcs status
Cluster name: mycluster
Stack: corosync
Current DC: pcmk-2 (version 1.1.18-11.el7_5.3-2b07d5c5a9) - partition with quorum
Last updated: Mon Sep 10 18:05:22 2018
Last change: Mon Sep 10 18:05:21 2018 by root via cibadmin on pcmk-1
2 nodes configured
5 resources configured
Online: [ pcmk-1 pcmk-2 ]
Full list of resources:
ClusterIP (ocf::heartbeat:IPaddr2): Started pcmk-2
WebSite (ocf::heartbeat:apache): Started pcmk-2
Master/Slave Set: WebDataClone [WebData]
Masters: [ pcmk-2 ]
Slaves: [ pcmk-1 ]
WebFS (ocf::heartbeat:Filesystem): Started pcmk-2
Daemon Status:
corosync: active/disabled
pacemaker: active/disabled
pcsd: active/enabled
Notice that pcmk-1 is back to the Online state, and that the cluster resources stay where they are due to our resource stickiness settings configured earlier.
Chapter 8. Configure STONITH
STONITH (Shoot The Other Node In The Head aka. fencing) protects your data from being corrupted by rogue nodes or unintended concurrent access.
Just because a node is unresponsive doesn’t mean it has stopped accessing your data. The only way to be 100% sure that your data is safe, is to use STONITH to ensure that the node is truly offline before allowing the data to be accessed from another node.
STONITH also has a role to play in the event that a clustered service cannot be stopped. In this case, the cluster uses STONITH to force the whole node offline, thereby making it safe to start the service elsewhere.
8.2. Choose a STONITH Device
It is crucial that your STONITH device can allow the cluster to differentiate between a node failure and a network failure.
A common mistake people make when choosing a STONITH device is to use a remote power switch (such as many on-board IPMI controllers) that shares power with the node it controls. If the power fails in such a case, the cluster cannot be sure whether the node is really offline, or active and suffering from a network fault, so the cluster will stop all resources to avoid a possible split-brain situation.
Likewise, any device that relies on the machine being active (such as SSH-based "devices" sometimes used during testing) is inappropriate.
For this example, assume we have a chassis containing four nodes and an IPMI device active on 10.0.0.1. Following the steps above would go something like this:
Step 1: Install the fence-agents-ipmilan package on both nodes.
Step 2: Configure the IP address, authentication credentials, etc. in the IPMI device itself.
Step 3: Choose the fence_ipmilan STONITH agent.
Step 4: Obtain the agent’s possible parameters:
[root@pcmk-1 ~]# pcs stonith describe fence_ipmilan
fence_ipmilan - Fence agent for IPMI
fence_ipmilan is an I/O Fencing agentwhich can be used with machines controlled by IPMI.This agent calls support software ipmitool (http://ipmitool.sf.net/). WARNING! This fence agent might report success before the node is powered off. You should use -m/method onoff if your fence device works correctly with that option.
Stonith options:
ipport: TCP/UDP port to use for connection with device
hexadecimal_kg: Hexadecimal-encoded Kg key for IPMIv2 authentication
port: IP address or hostname of fencing device (together with --port-as-ip)
inet6_only: Forces agent to use IPv6 addresses only
ipaddr: IP Address or Hostname
passwd_script: Script to retrieve password
method: Method to fence (onoff|cycle)
inet4_only: Forces agent to use IPv4 addresses only
passwd: Login password or passphrase
lanplus: Use Lanplus to improve security of connection
auth: IPMI Lan Auth type.
cipher: Ciphersuite to use (same as ipmitool -C parameter)
target: Bridge IPMI requests to the remote target address
privlvl: Privilege level on IPMI device
timeout: Timeout (sec) for IPMI operation
login: Login Name
verbose: Verbose mode
debug: Write debug information to given file
power_wait: Wait X seconds after issuing ON/OFF
login_timeout: Wait X seconds for cmd prompt after login
delay: Wait X seconds before fencing is started
power_timeout: Test X seconds for status change after ON/OFF
ipmitool_path: Path to ipmitool binary
shell_timeout: Wait X seconds for cmd prompt after issuing command
port_as_ip: Make "port/plug" to be an alias to IP address
retry_on: Count of attempts to retry power on
sudo: Use sudo (without password) when calling 3rd party sotfware.
priority: The priority of the stonith resource. Devices are tried in order of highest priority to lowest.
pcmk_host_map: A mapping of host names to ports numbers for devices that do not support host names. Eg. node1:1;node2:2,3 would tell the cluster to use port 1 for node1 and ports 2 and
3 for node2
pcmk_host_list: A list of machines controlled by this device (Optional unless pcmk_host_check=static-list).
pcmk_host_check: How to determine which machines are controlled by the device. Allowed values: dynamic-list (query the device), static-list (check the pcmk_host_list attribute), none
(assume every device can fence every machine)
pcmk_delay_max: Enable a random delay for stonith actions and specify the maximum of random delay. This prevents double fencing when using slow devices such as sbd. Use this to enable a
random delay for stonith actions. The overall delay is derived from this random delay value adding a static delay so that the sum is kept below the maximum delay.
pcmk_delay_base: Enable a base delay for stonith actions and specify base delay value. This prevents double fencing when different delays are configured on the nodes. Use this to enable
a static delay for stonith actions. The overall delay is derived from a random delay value adding this static delay so that the sum is kept below the maximum delay.
pcmk_action_limit: The maximum number of actions can be performed in parallel on this device Pengine property concurrent-fencing=true needs to be configured first. Then use this to
specify the maximum number of actions can be performed in parallel on this device. -1 is unlimited.
Default operations:
monitor: interval=60s
Step 5: pcs cluster cib stonith_cfg
Step 6: Here are example parameters for creating our STONITH resource:
[root@pcmk-1 ~]# pcs -f stonith_cfg stonith create ipmi-fencing fence_ipmilan \
pcmk_host_list="pcmk-1 pcmk-2" ipaddr=10.0.0.1 login=testuser \
passwd=acd123 op monitor interval=60s
[root@pcmk-1 ~]# pcs -f stonith_cfg stonith
ipmi-fencing (stonith:fence_ipmilan): Stopped
Steps 7-10: Enable STONITH in the cluster:
[root@pcmk-1 ~]# pcs -f stonith_cfg property set stonith-enabled=true
[root@pcmk-1 ~]# pcs -f stonith_cfg property
Cluster Properties:
cluster-infrastructure: corosync
cluster-name: mycluster
dc-version: 1.1.18-11.el7_5.3-2b07d5c5a9
have-watchdog: false
stonith-enabled: true
Step 11: pcs cluster cib-push stonith_cfg --config
Step 12: Test:
[root@pcmk-1 ~]# pcs cluster stop pcmk-2
[root@pcmk-1 ~]# stonith_admin --reboot pcmk-2
After a successful test, login to any rebooted nodes, and start the cluster (with pcs cluster start
).
Chapter 9. Convert Cluster to Active/Active
The primary requirement for an Active/Active cluster is that the data required for your services is available, simultaneously, on both machines. Pacemaker makes no requirement on how this is achieved; you could use a SAN if you had one available, but since DRBD supports multiple Primaries, we can continue to use it here.
9.1. Install Cluster Filesystem Software
The only hitch is that we need to use a cluster-aware filesystem. The one we used earlier with DRBD, xfs, is not one of those. Both OCFS2 and GFS2 are supported; here, we will use GFS2.
On both nodes, install the GFS2 command-line utilities and the Distributed Lock Manager (DLM) required by cluster filesystems:
# yum install -y gfs2-utils dlm
9.3. Create and Populate GFS2 Filesystem
Before we do anything to the existing partition, we need to make sure it is unmounted. We do this by telling the cluster to stop the WebFS resource. This will ensure that other resources (in our case, Apache) using WebFS are not only stopped, but stopped in the correct order.
[root@pcmk-1 ~]# pcs resource disable WebFS
[root@pcmk-1 ~]# pcs resource
ClusterIP (ocf::heartbeat:IPaddr2): Started pcmk-1
WebSite (ocf::heartbeat:apache): Stopped
Master/Slave Set: WebDataClone [WebData]
Masters: [ pcmk-1 ]
Slaves: [ pcmk-2 ]
WebFS (ocf::heartbeat:Filesystem): Stopped (disabled)
Clone Set: dlm-clone [dlm]
Started: [ pcmk-1 pcmk-2 ]
You can see that both Apache and WebFS have been stopped, and that pcmk-1 is the current master for the DRBD device.
Now we can create a new GFS2 filesystem on the DRBD device.
Warning
This will erase all previous content stored on the DRBD device. Ensure you have a copy of any important data.
Important
Run the next command on whichever node has the DRBD Primary role. Otherwise, you will receive the message:
/dev/drbd1: Read-only file system
[root@pcmk-1 ~]# mkfs.gfs2 -p lock_dlm -j 2 -t mycluster:web /dev/drbd1
It appears to contain an existing filesystem (xfs)
This will destroy any data on /dev/drbd1
Are you sure you want to proceed? [y/n] y
Discarding device contents (may take a while on large devices): Done
Adding journals: Done
Building resource groups: Done
Creating quota file: Done
Writing superblock and syncing: Done
Device: /dev/drbd1
Block size: 4096
Device size: 0.50 GB (131059 blocks)
Filesystem size: 0.50 GB (131056 blocks)
Journals: 2
Resource groups: 3
Locking protocol: "lock_dlm"
Lock table: "mycluster:web"
UUID: 0bcbffab-cada-4105-94d1-be8a26669ee0
The mkfs.gfs2
command required a number of additional parameters:
-p lock_dlm
specifies that we want to use the kernel’s DLM.
-j 2
indicates that the filesystem should reserve enough space for two journals (one for each node that will access the filesystem).
-t mycluster:web
specifies the lock table name. The format for this field is clustername:fsname
. For clustername
, we need to use the same value we specified originally with pcs cluster setup --name
(which is also the value of cluster_name in /etc/corosync/corosync.conf
). If you are unsure what your cluster name is, you can look in /etc/corosync/corosync.conf
or execute the command pcs cluster corosync pcmk-1 | grep cluster_name
.
Now we can (re-)populate the new filesystem with data (web pages). We’ll create yet another variation on our home page.
[root@pcmk-1 ~]# mount /dev/drbd1 /mnt
[root@pcmk-1 ~]# cat <<-END >/mnt/index.html
<html>
<body>My Test Site - GFS2</body>
</html>
END
[root@pcmk-1 ~]# chcon -R --reference=/var/www/html /mnt
[root@pcmk-1 ~]# umount /dev/drbd1
[root@pcmk-1 ~]# drbdadm verify wwwdata
9.5. Clone the IP address
There’s no point making the services active on both locations if we can’t reach them both, so let’s clone the IP address.
The IPaddr2 resource agent has built-in intelligence for when it is configured as a clone. It will utilize a multicast MAC address to have the local switch send the relevant packets to all nodes in the cluster, together with iptables clusterip rules on the nodes so that any given packet will be grabbed by exactly one node. This will give us a simple but effective form of load-balancing requests between our two nodes.
Let’s start a new config, and clone our IP:
[root@pcmk-1 ~]# pcs cluster cib loadbalance_cfg
[root@pcmk-1 ~]# pcs -f loadbalance_cfg resource clone ClusterIP \
clone-max=2 clone-node-max=2 globally-unique=true
clone-max=2
tells the resource agent to split packets this many ways. This should equal the number of nodes that can host the IP.
clone-node-max=2
says that one node can run up to 2 instances of the clone. This should also equal the number of nodes that can host the IP, so that if any node goes down, another node can take over the failed node’s "request bucket". Otherwise, requests intended for the failed node would be discarded.
globally-unique=true
tells the cluster that one clone isn’t identical to another (each handles a different "bucket"). This also tells the resource agent to insert iptables rules so each host only processes packets in its bucket(s).
Notice that when the ClusterIP becomes a clone, the constraints referencing ClusterIP now reference the clone. This is done automatically by pcs.
[root@pcmk-1 ~]# pcs -f loadbalance_cfg constraint
Location Constraints:
Ordering Constraints:
start ClusterIP-clone then start WebSite (kind:Mandatory)
promote WebDataClone then start WebFS (kind:Mandatory)
start WebFS then start WebSite (kind:Mandatory)
start dlm-clone then start WebFS (kind:Mandatory)
Colocation Constraints:
WebSite with ClusterIP-clone (score:INFINITY)
WebFS with WebDataClone (score:INFINITY) (with-rsc-role:Master)
WebSite with WebFS (score:INFINITY)
WebFS with dlm-clone (score:INFINITY)
Ticket Constraints:
Now we must tell the resource how to decide which requests are processed by which hosts. To do this, we specify the clusterip_hash parameter. The value of sourceip means that the source IP address of incoming packets will be hashed; each node will process a certain range of hashes.
[root@pcmk-1 ~]# pcs -f loadbalance_cfg resource update ClusterIP clusterip_hash=sourceip
Load our configuration to the cluster, and see how it responds.
[root@pcmk-1 ~]# pcs cluster cib-push loadbalance_cfg --config
CIB updated
[root@pcmk-1 ~]# pcs status
Cluster name: mycluster
Stack: corosync
Current DC: pcmk-1 (version 1.1.18-11.el7_5.3-2b07d5c5a9) - partition with quorum
Last updated: Tue Sep 11 10:36:38 2018
Last change: Tue Sep 11 10:36:33 2018 by root via cibadmin on pcmk-1
2 nodes configured
9 resources configured (1 DISABLED)
Online: [ pcmk-1 pcmk-2 ]
Full list of resources:
ipmi-fencing (stonith:fence_ipmilan): Started pcmk-1
WebSite (ocf::heartbeat:apache): Stopped
Master/Slave Set: WebDataClone [WebData]
Masters: [ pcmk-1 ]
Slaves: [ pcmk-2 ]
WebFS (ocf::heartbeat:Filesystem): Stopped (disabled)
Clone Set: dlm-clone [dlm]
Started: [ pcmk-1 pcmk-2 ]
Clone Set: ClusterIP-clone [ClusterIP] (unique)
ClusterIP:0 (ocf::heartbeat:IPaddr2): Started pcmk-2
ClusterIP:1 (ocf::heartbeat:IPaddr2): Started pcmk-1
Daemon Status:
corosync: active/disabled
pacemaker: active/disabled
pcsd: active/enabled
If desired, you can demonstrate that all request buckets are working by using a tool such as arping
from several source hosts to see which host responds to each.
9.6. Clone the Filesystem and Apache Resources
Now that we have a cluster filesystem ready to go, and our nodes can load-balance requests to a shared IP address, we can configure the cluster so both nodes mount the filesystem and respond to web requests.
Clone the filesystem and Apache resources in a new configuration. Notice how pcs automatically updates the relevant constraints again.
[root@pcmk-1 ~]# pcs cluster cib active_cfg
[root@pcmk-1 ~]# pcs -f active_cfg resource clone WebFS
[root@pcmk-1 ~]# pcs -f active_cfg resource clone WebSite
[root@pcmk-1 ~]# pcs -f active_cfg constraint
Location Constraints:
Ordering Constraints:
start ClusterIP-clone then start WebSite-clone (kind:Mandatory)
promote WebDataClone then start WebFS-clone (kind:Mandatory)
start WebFS-clone then start WebSite-clone (kind:Mandatory)
start dlm-clone then start WebFS-clone (kind:Mandatory)
Colocation Constraints:
WebSite-clone with ClusterIP-clone (score:INFINITY)
WebFS-clone with WebDataClone (score:INFINITY) (with-rsc-role:Master)
WebSite-clone with WebFS-clone (score:INFINITY)
WebFS-clone with dlm-clone (score:INFINITY)
Ticket Constraints:
Tell the cluster that it is now allowed to promote both instances to be DRBD Primary (aka. master).
[root@pcmk-1 ~]# pcs -f active_cfg resource update WebDataClone master-max=2
Finally, load our configuration to the cluster, and re-enable the WebFS resource (which we disabled earlier).
[root@pcmk-1 ~]# pcs cluster cib-push active_cfg --config
CIB updated
[root@pcmk-1 ~]# pcs resource enable WebFS
After all the processes are started, the status should look similar to this.
[root@pcmk-1 ~]# pcs resource
Master/Slave Set: WebDataClone [WebData]
Masters: [ pcmk-1 pcmk-2 ]
Clone Set: dlm-clone [dlm]
Started: [ pcmk-1 pcmk-2 ]
Clone Set: ClusterIP-clone [ClusterIP] (unique)
ClusterIP:0 (ocf::heartbeat:IPaddr2): Started pcmk-2
ClusterIP:1 (ocf::heartbeat:IPaddr2): Started pcmk-1
Clone Set: WebFS-clone [WebFS]
Started: [ pcmk-1 pcmk-2 ]
Clone Set: WebSite-clone [WebSite]
Started: [ pcmk-1 pcmk-2 ]
Testing failover is left as an exercise for the reader. For example, you can put one node into standby mode, use pcs status
to confirm that its ClusterIP clone was moved to the other node, and use arping
to verify that packets are not being lost from any source host.
Note
You may find that when a failed node rejoins the cluster, both ClusterIP clones stay on one node, due to the resource stickiness. While this works fine, it effectively eliminates load-balancing and returns the cluster to an active-passive setup again. You can avoid this by disabling stickiness for the IP address resource:
[root@pcmk-1 ~]# pcs resource meta ClusterIP resource-stickiness=0