[클러스터] GFS+CLVM installation document

How to install and run clvm and gfs.

Refer to the cluster project page for the latest information.

http://sources.redhat.com/cluster/

Get source

———-

– download the source tarballs

  latest linux kernel        – ftp://ftp.kernel.org/pub/linux/

  device-mapper                – ftp://sources.redhat.com/pub/dm/

  lvm2                        – ftp://sources.redhat.com/pub/lvm2/

  iddev                        – ftp://sources.redhat.com/pub/cluster/

  ccs                        – ftp://sources.redhat.com/pub/cluster/

  fence                        – ftp://sources.redhat.com/pub/cluster/

  cman                        – ftp://sources.redhat.com/pub/cluster/

  cman-kernel                – ftp://sources.redhat.com/pub/cluster/

  dlm                        – ftp://sources.redhat.com/pub/cluster/

  dlm-kernel                – ftp://sources.redhat.com/pub/cluster/

  gfs                        – ftp://sources.redhat.com/pub/cluster/

  gfs-kernel                – ftp://sources.redhat.com/pub/cluster/

– or to download source from cvs see

  http://sources.redhat.com/dm/

  http://sources.redhat.com/lvm2/

  http://sources.redhat.com/cluster/

  summary: after cvs login,

  cvs -d :pserver:cvs@sources.redhat.com:/cvs/dm      checkout device-mapper

  cvs -d :pserver:cvs@sources.redhat.com:/cvs/lvm2    checkout LVM2

  cvs -d :pserver:cvs@sources.redhat.com:/cvs/cluster checkout cluster

Build and install

—————–

– apply kernel patches from

  cman-kernel/patches/

  dlm-kernel/patches/

  gfs-kernel/patches/

  

– configure kernel, selecting DM, CMAN, DLM, GFS

  (the final three, at least, should be built as modules)

– build and install kernel and modules

– build and install userland programs and libraries (order is important)

  device-mapper

  ./configure

  make; make install

  lvm2

  ./configure –with-clvmd –with-cluster=shared

  make; make install

  scripts/clvmd_fix_conf.sh /lib/liblvm2clusterlock.so

  ccs

  ./configure –kernel_src=/path/to/patched/kernel

  make; make install

  cman

  ./configure –kernel_src=/path/to/patched/kernel

  make; make install

  dlm

  ./configure –kernel_src=/path/to/patched/kernel

  make; make install

  fence

  ./configure –kernel_src=/path/to/patched/kernel

  make; make install

  iddev

  ./configure

  make; make install

  gfs

  ./configure –kernel_src=/path/to/patched/kernel

  make; make install

Load kernel modules

——————-

depmod -a

modprobe dm-mod

device-mapper/scripts/devmap_mknod.sh

modprobe gfs

modprobe lock_dlm

Modules that should be loaded: lock_dlm, dlm, cman, gfs, lock_harness

and dm-mod if device-mapper was built as a module

Startup procedure

—————–

Run these commands on each cluster node:

> ccsd                             – Starts the CCS daemon

> cman_tool join                   – Joins the cluster

> fence_tool join                  – Joins the fence domain (starts fenced)

> clvmd                            – Starts the CLVM daemon

> vgchange -aly                    – Activates LVM volumes (locally)

> mount -t gfs /dev/vg/lvol /mnt   – Mounts a GFS file system

Shutdown procedure

——————

Run these commands on each cluster node:

> umount /mnt                           – Unmounts a GFS file system

> vgchange -aln                    – Deactivates LVM volumes (locally)

> killall clvmd                    – Stops the CLVM daemon

> fence_tool leave                 – Leaves the fence domain (stops fenced)

> cman_tool leave                  – Leaves the cluster

> killall ccsd                     – Stops the CCS daemon

Creating CCS config

——————-

There is no GUI or command line program to create the config file yet.

The cluster config file “cluster.xml” must therefore be created manually.

Once created, cluster.xml should be placed in the /etc/cluster/ directory

on one cluster node.  CCS daemon (ccsd) will take care of transferring it

to other nodes where it’s needed.  (FIXME: updating cluster.xml in a running

cluster is supported but not documented.)

A minimal cluster.xml example is shown below.

Creating CLVM logical volumes

—————————–

Use standard LVM commands (see LVM documentation on using pvcreate, vgcreate,

lvcreate.)  A node must be running the CLVM system to use the LVM commands.

Running the CLVM system means successfully running the commands above up

through starting clvmd.

Creating GFS file systems

————————-

> gfs_mkfs -p lock_dlm -t <ClusterName>:<FSName> -j <Journals> <Device>

        <ClusterName> must match the cluster name used in CCS config

        <FSName> is a unique name chosen now to distinguish this fs from others

        <Journals> the number of journals in the fs, one for each node to mount

        <Device> a block device, usually an LVM logical volume

Creating a GFS file system means writing to a CLVM volume which means the CLVM

system must be running (see previous section.)

Cluster startup/shutdown notes

——————————

Fencing: In the start-up steps above, “fence_tool join” is the equivalent of

simply starting fenced.  fence_tool is useful because additional options can be

specified to delay the actual starting of fenced.  Delaying can be useful to

avoid unnecessarily fencing nodes that haven’t joined the cluster yet.  The

only option fence_tool now provides to address this is “-t <seconds>” to wait

the given number of seconds before starting fenced.

Shutdown: There is also a practical timing issue with respect to the shutdown

steps being run on all nodes when shutting down an entire cluster.  When

shutting down the entire cluster (or shutting down a node for an extended

period) use “cman_tool leave remove”.  This automatically reduces the number of

expected votes as each node leaves and prevents the loss of quorum which could

keep the last nodes from cleanly completing shutdown.

Using the “remove” leave option should not be used in general since it

introduces potential split-brain risks.

If the “remove” leave option is not used, quorum will be lost after enough

nodes have left the cluster.  Once the cluster is inquorate, remaining members

that have not yet completed “fence_tool leave” in the steps above will be

stuck.  Operations such as umounting gfs or leaving the fence domain

(“fence_tool leave”) will block while the cluster is inquorate.  They can

continue and complete only once quorum is regained.

If this happens, one option is to join the cluster (“cman_tool join”) on some

of the nodes that have left so that the cluster regains quorum and the stuck

nodes can complete their shutdown.  Another option is to forcibly reduce the

number of expected votes for the cluster which allows the cluster to become

quorate again (“cman_tool expected <votes>”).  This later method is the

equivalent to using the “remove” option when leaving.

Config file

———–

This example primarily illustrates the variety of fencing configurations.

The first node uses “cascade fencing”; if the first method fails (power cycling

with an APC Masterswitch), the second is tried (port disable on a Brocade FC

switch).  In this example, the node has dual paths to the storage so the port

on both paths must be disabled (the same idea applies to nodes with dual power

supplies.)

There is only one method of fencing the second node (via an APC Masterswitch)

so no cascade fencing is possible.

If no hardware is available for fencing, manual fencing can be used as shown

for the third node.  If a node with manual fencing fails, a human must take

notice (a message appears in the system log) and run fence_ack_manual after

resetting the failed node.  (The node that actually carries out fencing

operations is the node with the lowest ID in the fence domain.)

<?xml version=”1.0″?>

<cluster name=”alpha” config_version=”1″>

<cman>

</cman>

<nodes>

<node name=”nd01″ votes=”1″>

        <fence>

                <method name=”cascade1″>

                        <device name=”apc1″ port=”1″/>

                </method>

                <method name=”cascade2″>

                        <device name=”brocade1″ port=”1″/>

                        <device name=”brocade2″ port=”1″/>

                </method>

        </fence>

</node>

<node name=”nd02″ votes=”1″>

        <fence>

                <method name=”single”>

                        <device name=”apc1″ port=”2″/>

                </method>

        </fence>

</node>

<node name=”nd03″ votes=”1″>

        <fence>

                <method name=”single”>

                        <device name=”human” ipaddr=”nd03″/>

                </method>

        </fence>

</node>

</nodes>

<fence_devices>

        <device name=”apc1″ agent=”fence_apc” ipaddr=”10.1.1.1″ login=”apc” passwd=”apc”/>

        <device name=”brocade1″ agent=”fence_brocade” ipaddr=”10.1.1.2″ login=”user” passwd=”pw”/>

        <device name=”brocade2″ agent=”fence_brocade” ipaddr=”10.1.1.3″ login=”user” passwd=”pw”/>

        <device name=”human” agent=”fence_manual”/>

</fence_devices>

</cluster>

Multiple clusters

—————–

When multiple clusters are used, it can be useful to specify the cluster name

on the cman_tool command line.  This forces CCS to select a cluster.xml with

the same cluster name.  The node then joins this cluster.

> cman_tool join -c <ClusterName>

[Note: If the -c option is not used, ccsd will first check the local copy of

cluster.xml to extract the cluster name and will only grab a remote copy of

cluster.xml if it has the same cluster name and a greater version number.  If a

local copy of cluster.xml does not exist, ccsd may grab a cluster.xml for a

different cluster than intended — cman_tool would then report an error that

the node is not listed in the file.

So, if you don’t currently have a local copy of cluster.xml (and there are

other clusters running) or you wish to join a different cluster with a

different cluster.xml from what exists locally, you must specify the -c

option.]

Two node clusters

—————–

Ordinarily the loss of quorum after one node fails out of two will prevent the

remaining node from continuing (if both nodes have one vote.)  Some special

configuration options can be set to allow the one remaining node to continue

operating if the other fails.  To do this only two nodes with one vote each can

be defined in cluster.xml.  The two_node and expected_votes values must then be

set to 1 in the cman config section as follows.

  <cman two_node=”1″ expected_votes=”1″>

  </cman>

Advanced Network Configuration

——————————

* Multihome

CMAN can be configured to use multiple network interfaces.  If one fails it

should be able to continue running with the one remaining.  A node’s name in

cluster.xml is always associated with the IP address on one network interface;

“nd1” in the following:

<node name=”nd1″ votes=”1″>

</node>

To use a second network interface, the node must have a second hostname

associated with the IP address on that interface; “nd1-e1” in the following.

The second hostname is specfied in an “altname” section.

<node name=”nd1″ votes=”1″>

    <altname name=”nd1-e1″/>

</node>

* Multicast

CMAN can be configured to use multicast instead of broadcast (broadcast is used

by default if no multicast parameters are given.)  To configure multicast when

one network interface is used add one line under the <cman> section and another

under the <node> section:

<cman>

    <multicast addr=”224.0.0.1″/>

</cman>

<node name=”nd1″ votes=”1″>

    <multicast addr=”224.0.0.1″ interface=”eth0″/>

</node>

The multicast addresses must match and the address must be usable on the

interface name given for the node.

When two interfaces are used, multicast is configured as follows:

<cman>

    <multicast addr=”224.0.0.1″/>

    <multicast addr=”224.0.0.9″/>

</cman>

<node name=”nd1″ votes=”1″>

    <altname name=”nd1-e1″/>

    <multicast addr=”224.0.0.1″ interface=”eth0″/>

    <multicast addr=”224.0.0.9″ interface=”eth1″/>

</node>

* IPv6

– When using multiple interfaces, all must use the same address family.  Mixing

  IPv4 and IPv6 is not allowed.

– When using IPv6, multicast must be configured; there is no IPv6 broadcast.

서진우

서진우

슈퍼컴퓨팅 전문 기업 클루닉스/ 상무(기술이사)/ 정보시스템감리사/ 시스존 블로그 운영자

You may also like...

페이스북/트위트/구글 계정으로 댓글 가능합니다.