Service Card for NFS

• ha nfs setup:
  

/usr/lib64/heartbeat/hb_takeover foreign

once you think everything is in good order again, go to the original server and take back the resources:

/usr/lib64/heartbeat/hb_takeover local

May 30 14:59 [root@nfs02:~]# cat /etc/ha.d/haresources 
nfs01.lcg.cscs.ch IPaddr::148.187.67.100 \
drbddisk::srv_nfs \
Filesystem::/dev/drbd0::/srv/nfs::ext3 \
drbddisk::export \
Filesystem::/dev/drbd2::/export::ext3 \
drbddisk::home \
Filesystem::/dev/drbd3::/home::ext3 \
drbddisk::atlas \
Filesystem::/dev/drbd4::/experiment_software/atlas::ext3 \
drbddisk::cms \
Filesystem::/dev/drbd5::/experiment_software/cms::ext3 \
drbddisk::lhcb \
Filesystem::/dev/drbd6::/experiment_software/lhcb::ext3 \
drbddisk::other \
Filesystem::/dev/drbd7::/experiment_software/others::ext3 \
drbddisk::shared \
Filesystem::/dev/drbd8::/shared::ext3 nfslock nfs httpd

nfs02.lcg.cscs.ch IPaddr::148.187.67.109 \
drbddisk::srv_cfengine \
Filesystem::/dev/drbd1::/srv/cfengine::ext3 cfservd

May 30 15:00 [root@nfs02:~]# cat /etc/ha.d/ha.cf 
logfacility     local0
debugfile /var/log/ha-debug
logfile /var/log/ha-log

keepalive 2
deadtime 30
initdead 120

udpport 694
#
bcast eth0
bcast eth3

auto_failback off

#       Tell what machines are in the cluster
#       node    nodename ...    -- must match uname -n
node    nfs01.lcg.cscs.ch nfs02.lcg.cscs.ch

crm no

mdadm --create /dev/md0 -v --raid-devices=4 --chunk=32 --level=raid10 /dev/sdc /dev/sdd /dev/sde /dev/sdf
pvcreate /dev/md0

vgcreate vgnfs /dev/md0
lvcreate --size 1G --name lv_srv_nfs vgnfs
lvcreate --size 1G --name lv_srv_cfengine vgnfs
lvcreate --size 5G --name lv_export vgnfs
lvcreate --size 100G --name lv_home vgnfs
lvcreate --size 500G --name lv_atlas vgnfs
lvcreate --size 300G --name lv_cms vgnfs
lvcreate --size 5G --name drbd_meta_disk vgnfs
lvcreate --size 200G --name lv_lhcb vgnfs
lvcreate --size 1G --name lv_other_vo vgnfs
lvcreate --size 20G --name lv_shared vgnfs
rpm -Uvh http://download.fedora.redhat.com/pub/epel/5/x86_64/epel-release-5-4.noarch.rpm
yum --enablerepo=epel -y install heartbeat.x86_64

modprobe drbd
chkconfig nfs off
drbdadm create-md srv_nfs
echo yes | drbdadm create-md srv_cfengine
echo yes | drbdadm create-md export
echo yes | drbdadm create-md home
echo yes | drbdadm create-md atlas
echo yes | drbdadm create-md cms
echo yes | drbdadm create-md lhcb
echo yes | drbdadm create-md other
echo yes | drbdadm create-md shared
#drbdadm up all
#service drbd status
mkdir /export
mkdir -p /srv/cfengine
mkdir -p /srv/nfs
mkdir -p /experiment_software/{atlas,cms,lhcb,others}
mkdir -p /shared

NAS appliance

Phoenix has access to part to an EMC NAS appliance installed at CSCS. Currently the quota available for all mount points is ~100TB and consists of 6 heads with several 10GbE links towards the cluster.

The following filesystems are exported from this appliance:

• /ifs/LCG/backup is intended to replace the old backup system on the shared fs of /project
• /ifs/LCG/cm holds the RPM repositories and all the configuration management software required to make Phoenix run
• /ifs/LCG/experiment_software is intended to be used as a central point where VOs will put their software. Currently only ATLAS requires minimal space here.
• /ifs/LCG/home this is the home of all users that have access to Phoenix.
• /ifs/LCG/shared Some important subdirectories are stored here (such as apelaccounting, exp_soft_arc, slurm, gridmapdir, etc.)
• /ifs/LCG/syslog Where all logs are stored.
• /ifs/LCG/vm This space holds the VMs that form the preproduction cluster.

-- JasonTemple - 2011-05-30