ZFS Networker

Backup of ZFS snapshots on Solaris Cluster with Legato/EMC Networker

This describes how to setup a backup of the Solaris Cluster resource group named sample-rg.

The structure of my RGs is always:

RG:                <name>-rg
ZFS-HASP:          <name>-hasp-zfs-res
Logical Host:      <name>-lh-res
Logical Host Name: <name>-lh
ZPOOL:             <name>_pool

I used the bash as shell.

Define variables used in the following command lines

# NAME=sample
# RGname=${NAME}-rg
# NetworkerGroup=$(echo ${NAME} | tr 'a-z' 'A-Z' )
# ZPOOL=${NAME}_pool
# ZPOOL_BASEDIR=/local/${RGname}

Define a resource for Networker

What we need now is a resource definition in our Networker directory like this:

# zfs create ${ZPOOL}/nsr
# mkdir ${ZPOOL_BASEDIR}/nsr/{bin,log,res}

# cat > ${ZPOOL_BASEDIR}/nsr/res/${NetworkerGroup}.res <<EOF
type: savepnpc;
precmd: "${ZPOOL_BASEDIR}/nsr/bin/prepst_command.sh pre >${ZPOOL_BASEDIR}/nsr/log/networker_precmd.log 2>&1";
pstcmd: "${ZPOOL_BASEDIR}/nsr/bin/prepst_command.sh pst >${ZPOOL_BASEDIR}/nsr/log/networker_pstcmd.log 2>&1";
timeout: "08:00am";
abort precmd with group: Yes;
EOF

And now create a link on every cluster node to this file

# ln -s ${ZPOOL_BASEDIR}/nsr/res/${NetworkerGroup}.res /nsr/res/${NetworkerGroup}.res

The pre-/pstcmd-script

!!!THIS CODE IS UNTESTED DO NOT USE THIS!!!

!!!THIS JUST AN EXAMPLE!!!

#!/bin/bash
ZPOOL="sample_pool"
ZONE="sample-zone"
ORACLE_SID=SAMPLE
ORACLE_USER=oracle

SNAPSHOT_NAME="nsr"
ZFS_CMD="echo /usr/bin/zfs"
ZLOGIN_CMD="echo /usr/bin/zlogin"

function snapshot_pre {
  DB=$1
  DBUSER=$2
  if [ $# -eq 3 && "_$3_" != "__" ]
  then 
    ZONE=$3
    ZONE_CMD="${ZLOGIN_CMD} -l ${DBUSER} ${ZONE}"
    ZONE_BASE=$(/usr/sbin/zonecfg -z ${ZONE} info zonepath | nawk '{print $NF;}')
    ZONE_ROOT="${ZONE_BASE}/root"
  else
    ZONE_ROOT=""
  fi
  SCRIPT_NAME="tmp/.nsr-pre-snap-script.$$"

  # Create script inside zone
  cat >${ZONE_ROOT}/{SCRIPT_NAME} <<EOS
#!/bin/bash
DBDIR=\$(/usr/bin/nawk -F':' -v ORACLE_SID=${ORACLE_SID} '\$1==ORACLE_SID {print \$2;}' /var/opt/oracle/oratab)
\${DBDIR}/bin/sqlplus sys/${DBUSER} as sysdba << EOF
create pfile from spfile;
alter system archive log current;
alter database backup controlfile to trace;
alter database begin backup;
EOF
EOS
  chmod 755 ${ZONE_ROOT}/${SCRIPT_NAME}

  ${ZONE_CMD} /${SCRIPT_NAME} 2>&1 | print_log
  rm -f ${ZONE_ROOT}/${SCRIPT_NAME}
}

function snapshot_pst {
  DB=$1
  DBUSER=$2
  if [ $# -eq 3 && "_$3_" != "__" ]
  then
    ZONE=$3
    ZONE_CMD="${ZLOGIN_CMD} -l ${DBUSER} ${ZONE}"
    ZONE_BASE=$(/usr/sbin/zonecfg -z ${ZONE} info zonepath | nawk '{print $NF;}')
    ZONE_ROOT="${ZONE_BASE}/root"
  else
    ZONE_ROOT=""
  fi
  SCRIPT_NAME="tmp/.nsr-pre-snap-script.$$"

  # Create script inside zone
  cat >${ZONE_ROOT}/{SCRIPT_NAME} <<EOS
#!/bin/bash
DBDIR=\$(/usr/bin/nawk -F':' -v ORACLE_SID=${ORACLE_SID} '\$1==ORACLE_SID {print \$2;}' /var/opt/oracle/oratab)
\${DBDIR}/bin/sqlplus sys/${DBUSER} as sysdba << EOF
alter database end backup;
alter system archive log current;
EOF
EOS
  chmod 755 ${ZONE_ROOT}/${SCRIPT_NAME}

  ${ZONE_CMD} /${SCRIPT_NAME} 2>&1 | print_log
  rm -f ${ZONE_ROOT}/${SCRIPT_NAME}
}

function snapshot_create {
  ZPOOL=$1
  SNAPSHOT_NAME=$2 
  ${ZFS_CMD} snapshot -r ${ZPOOL}@${SNAPSHOT_NAME} 2>&1 | print_log
}

function snapshot_destroy {
  ZPOOL=$1
  SNAPSHOT_NAME=$2 
  if (${ZFS_CMD} list -t snapshot ${ZPOOL}@${SNAPSHOT_NAME} >/dev/null)
  then
    ${ZFS_CMD} destroy -r ${ZPOOL}@${SNAPSHOT_NAME} 2>&1 | print_log
  fi
}

function usage {
  echo "Usage: $0 (pre|pst)"
  exit 1
}

case $1 in
pre)
  snapshot_destroy ${ZPOOL} ${SNAPSHOT_NAME}
  snapshot_pre     ${DB}    ${DBUSER}        ${ZONE}
  snapshot_create  ${ZPOOL} ${SNAPSHOT_NAME}
  snapshot_pst     ${DB}    ${DBUSER}        ${ZONE}
  ;;
pst)
  snapshot_destroy ${ZPOOL} ${SNAPSHOT_NAME}
  ;;
*)
  usage 
  ;;
esac

!!!THIS CODE IS UNTESTED DO NOT USE THIS!!!

!!!THIS JUST AN EXAMPLE!!!

Registering new resource type LGTO.clnt

1. Install Solaris client package LGTOclnt 2. Register new resource type in cluster. One one node do:

# clrt register -f /usr/sbin/LGTO.clnt.rtr LGTO.clnt

Now you have a new resource type LGTO.clnt in your cluster.

Create client resource of type LGTO.clnt

So I use scripts like this:

# RGname=sample-rg
# clrs create \
  -t LGTO.clnt \
  -g ${RGname} \
  -p Resource_dependencies=$(basename ${RGname} -rg)-hasp-zfs-res \
  -p clientname=$(basename ${RGname} -rg)-lh \
  -p Network_resource=$(basename ${RGname} -rg)-lh-res \
  -p owned_paths=${ZPOOL_BASEDIR} \
  $(basename ${RGname} -rg)-nsr-res

This expands to:

# clrs create \
  -t LGTO.clnt \
  -g sample-rg \
  -p Resource_dependencies=sample-hasp-zfs-res \
  -p clientname=sample-lh \
  -p Network_resource=sample-lh-res \
  -p owned_paths=/local/sample-rg \
  sample-nsr-res

Now we have a client name to which we can connect to: sample-lh