load_checkpoint.sh #3

#!/bin/bash
#------------------------------------------------------------------------------
### WARNING WARNING WARNING WARNING WARNING WARNING WARNING WARNING WARNING
### THIS SCRIPT SHOULD BE USED WITH GREAT CAUTION.
#------------------------------------------------------------------------------
set -u

# This script loads a specified checkpoint simultaneously into /p4/N/root
# and /p4/N/offline_db.

#==============================================================================
# Declarations and Environment

# Allow override of P4U_HOME, which is set only when testing P4U scripts.
export P4U_HOME="${P4U_HOME:-/p4/common/bin}"
export P4U_LIB="${P4U_LIB:-/p4/common/lib}"
export P4U_ENV="$P4U_LIB/p4u_env.sh"
export P4U_LOG=Unset
export VERBOSITY=${VERBOSITY:-5}

# Environment isolation.  For stability and security reasons, prepend
# PATH to include dirs where known-good scripts exist.
# known/tested PATH and, by implication, executables on the PATH.
export PATH="$P4U_HOME:$PATH:~/bin:."
export P4CONFIG="${P4CONFIG:-.p4config}"

[[ -r "$P4U_ENV" ]] || {
   echo -e "\\nError: Cannot load environment from: $P4U_ENV\\n\\n"
   exit 1
}

declare BASH_LIBS=$P4U_ENV
BASH_LIBS+=" $P4U_LIB/libcore.sh"
BASH_LIBS+=" $P4U_LIB/libp4u.sh"

for bash_lib in $BASH_LIBS; do
   # shellcheck disable=SC1090
   source "$bash_lib" ||\
      { echo -e "\\nFATAL: Failed to load bash lib [$bash_lib]. Aborting.\\n"; exit 1; }
done

declare Version=2.1.1
declare -i SilentMode=0
declare Checkpoint=Unset
declare SDPInstance=
declare SDPInstanceVarsFile=
declare OfflineDB=
declare OfflineDBUsableFile=
declare CaseMode=
declare CaseFlag=
declare CompressFlag=
declare ExcludedTables=
declare P4DInitScript=
declare P4DSystemdServiceFile=
declare P4BrokerInitScript=
declare P4BrokerSystemdServiceFile=
declare P4BrokerCfg=
declare P4DPids=
declare ThisHost=${HOSTNAME%%.*}

#==============================================================================
# Local Functions

function user_confirmation_and_warning () {
   local interactive=${1:-1}
   local input=""
   msg "\\n
==============================================================================
WARNING WARNING WARNING WARNING WARNING WARNING WARNING WARNING WARNING\\n

This script will remove files in P4ROOT and the offline_db folder,
including db.* files, state* files, and others.

It is STRONGLY recommended that these files be preseved first unless
it is absolutely certain they will never potentially be useful in any
recovery scenario.

This is being run on host $HOSTNAME at $(date).

WARNING WARNING WARNING WARNING WARNING WARNING WARNING WARNING WARNING
==============================================================================
\\n\\n"

   if [[ $interactive -eq 0 ]]; then
      msg "Bypassing interactive confirmation due to -y on command line."
      return 0
   fi

   while [[ -z "$input" ]]; do
      echo -e -n "\\nEnter Y to proceed, N to stop [y/Y/n/N]:  "
      read -r -e input
      if [[ "${input^^}" == "Y" ]]; then
         continue
      elif [[ "${input^^}" == "N" ]]; then
         msg "Confirmation to proceed not received."
         return 1
      else
         input=
      fi
   done

   return 0
}


#------------------------------------------------------------------------------
# Function: terminate
function terminate
{
   # Disable signal trapping.
   trap - EXIT SIGINT SIGTERM

   vvmsg "$THISSCRIPT: EXITCODE: $OverallReturnStatus"

   # Stop logging.
   [[ "${P4U_LOG}" == off ]] || stoplog

   # Don't litter.
   cleanTrash

   # With the trap removed, exit.
   exit "$OverallReturnStatus"
}

#------------------------------------------------------------------------------
# Function: usage (required function)
#
# Input:
# $1 - style, either -h (for short form) or -man (for man-page like format).
# The default is -h.
#
# $2 - error message (optional).  Specify this if usage() is called due to
# user error, in which case the given message displayed first, followed by the
# standard usage message (short or long depending on $1).  If displaying an
# errror, usually $1 should be -h so that the longer usage message doesn't
# obsure the error message.
#
# Sample Usage:
# usage
# usage -h
# usage -man
# usage -h "Incorrect command line usage."
#------------------------------------------------------------------------------
function usage
{
   declare style=${1:--h}
   declare errorMessage=${2:-Unset}

   if [[ "$errorMessage" != Unset ]]; then
      echo -e "\\n\\nUsage Error:\\n\\n$errorMessage\\n\\n"
   fi

   echo "USAGE for $THISSCRIPT v$Version:

$THISSCRIPT <checkpoint> [-i <instance>] [-r] [-b] [-y] [-L <log>] [-si] [-v<n>] [-D]

or

$THISSCRIPT [-h|-man|-V]
"
   if [[ $style == -man ]]; then
      echo -e "
DESCRIPTION:
	This script loads a specified checkpoint into /p4/N/root and /p4/N/offline_db.

ARGUMENTS AND OPTIONS:
 <checkpoint>
	Specify the path to the checkpoint file to load.

	The file may be a compressed or uncompressed checkpoint.

	This positional paremter is required.

 -i <instance>
	Specify the SDP instance. This can be omitted if SDP_INSTANCE is already
	defined.

 -r	Specify '-r' to replay only to P4ROOT.  By default, this script replays both
	to P4ROOT and the offline_db.

 -b	Specify '-b' to start the a p4broker process (if configured).  By default
	the p4d process is started after loading the checkpoint, but the p4broker
	process is not. This can be useful to ensure the human administrator has
	an opportunity to do sanity checks before enabling the broker to allow
	access by end users (if the broker is deployed for this usage).

 -y	Use the '-y' flag to bypass an interactive warning and confirmation
	prompt.

 -v<n>	Set verbosity 1-5 (-v1 = quiet, -v5 = highest).  The default is 5.

 -L <log>
	Specify the path to a log file, or the special value 'off' to disable
	logging.  By default, all output (stdout and stderr) goes to:
	${LOGS:-/tmp}/${THISSCRIPT%.sh}.<timestamp>.log

	NOTE: This script is self-logging.  That is, output displayed on the screen
	is simultaneously captured in the log file.  Do not run this script with
	redirection operators like '> log' or '2>&1', and do not use 'tee.'

 -si	Operate silently.  All output (stdout and stderr) is redirected to the log
	only; no output appears on the terminal.  This cannot be used with
	'-L off'.
      
 -D     Set extreme debugging verbosity.

HELP OPTIONS:
 -h	Display short help message
 -man	Display man-style help message
 -V	Dispay version info for this script and its libraries.

FILES:

EXAMPLES:
	Sample Usage (bash syntax):
	nohup $P4CBIN/load_checkpoint.sh /p4/1/checkpoints/p4_1.ckp.4025.gz -i 1 -y -si < /dev/null > /dev/null 2>&1 &

	Then, monitor with:
	tail -f \$(ls -t \$LOGS/load_checkpoint.*.log|head -1)\\n
"
   fi

   exit 1
}

#==============================================================================
# Command Line Processing

declare -i LoadOfflineDB=1
declare -i StartBrokerWhenDone=0
declare -i Interactive=1
declare -i shiftArgs=0

set +u
while [[ $# -gt 0 ]]; do
   case $1 in
      (-r) LoadOfflineDB=0;;
      (-b) StartBrokerWhenDone=1;;
      (-y) Interactive=0;;
      (-i) SDPInstance="$2"; shiftArgs=1;;
      (-h) usage -h;;
      (-man) usage -man;;
      (-V) show_versions; exit 1;;
      (-v1) export VERBOSITY=1;;
      (-v2) export VERBOSITY=2;;
      (-v3) export VERBOSITY=3;;
      (-v4) export VERBOSITY=4;;
      (-v5) export VERBOSITY=5;;
      (-L) export u4U_LOG="$2"; shiftArgs=1;;
      (-si) SilentMode=1;;
      (-D) set -x;; # Debug; use 'set -x' mode.
      (-*) usage -h "Unknown flag ($1).";;
      (*)
         if [[ "$Checkpoint" == Unset ]]; then
            Checkpoint="$1"
         else
            usage -h "Only one checkpoint file can be specified."
         fi
      ;;
   esac

   # Shift (modify $#) the appropriate number of times.
   shift; while [[ $shiftArgs -gt 0 ]]; do
      [[ $# -eq 0 ]] && usage -h "Incorrect number of arguments."
      shiftArgs=$shiftArgs-1
      shift
   done
done
set -u

#==============================================================================
# Command Line Verification

[[ "$SilentMode" -eq 1 && "$P4U_LOG" == off ]] && \
   usage -h "Cannot use '-si' with '-L off'."

[[ "${P4U_LOG:-Unset}" == Unset ]] && \
   P4U_LOG="${LOGS:-/tmp}/${THISSCRIPT%.sh}.$(date +'%Y%m%d-%H%M%S').log"

if [[ -z "$SDPInstance" ]]; then
   if [[ "${SDP_INSTANCE:-Unset}" != Unset ]]; then
      SDPInstance="$SDP_INSTANCE"
   else
      usage -h "SDP Instance parameter is missing and SDP_INSTANCE is undefined." 
   fi
fi

#==============================================================================
# Main Program

trap terminate EXIT SIGINT SIGTERM

declare -i OverallReturnStatus=0

if [[ "${P4U_LOG}" != off ]]; then
   touch "${P4U_LOG}" || bail "Couldn't touch log file [${P4U_LOG}]."

   # Redirect stdout and stderr to a log file.
   if [[ "$SilentMode" -eq 0 ]]; then
      exec > >(tee "${P4U_LOG}")
      exec 2>&1
   else
      exec >"${P4U_LOG}"
      exec 2>&1
   fi

   initlog
fi

msg "Started ${0##*/} v$Version on $ThisHost at $(date)"

# shellcheck disable=SC2153
P4DInitScript="$P4HOME/bin/p4d_${SDPInstance}_init"
P4DSystemdServiceFile="/etc/systemd/system/p4d_${SDPInstance}.service"
P4BrokerInitScript="$P4HOME/bin/p4broker_${SDPInstance}_init"
P4BrokerSystemdServiceFile="/etc/systemd/system/p4broker_${SDPInstance}.service"
SDPInstanceVarsFile="/p4/common/config/p4_${SDPInstance}.vars"

[[ "$Checkpoint" == Unset ]] && usageError "Missing checkpoint parameter."
[[ -r "$SDPInstanceVarsFile" ]] ||\
   bail "Missing SDP instance vars file. Typo in instance name? Expected: $SDPInstanceVarsFile"
[[ -r "$Checkpoint" ]] ||\
   bail "Specified checkpoint file does not exist: $Checkpoint"

# shellcheck disable=SC1091
source /p4/common/bin/p4_vars "$SDPInstance"
# shellcheck disable=SC1091
source /p4/common/bin/ps_functions.sh ||\
   bail "Failed to load ps_functions.sh."
#shellcheck disable=SC1091
source /p4/common/bin/edge_vars ||\
   bail "Failed to load edge_vars."
declare CheckpointMD5=

[[ -n "$ExcludedTables" ]] ||\
   bail "Value for \$ExludedTables not defined in $P4CBIN/edge_vars."

user_confirmation_and_warning "$Interactive" ||\
   bail "User confirmation not verified. Aborting."

cd "$P4ROOT" || bail "Could not cd to P4ROOT [$P4ROOT]."

[[ "$PWD" == "/p4/${SDPInstance}/root" ]] || bail "Unexpected P4ROOT value of: $P4ROOT"

CheckpointMD5="${Checkpoint}.md5"
if [[ -r "$CheckpointMD5" ]]; then
   msg "Verified: MD5 file for checkpoint exists: $CheckpointMD5"
else
   CheckpointMD5="${Checkpoint/.gz/.md5}"
   if [[ -r "$CheckpointMD5" ]]; then
      msg "Verified: MD5 file for checkpoint exists: $CheckpointMD5"
   else
      bail "Could not find MD5 file for checkpoint. Tried ${Checkpoint}.md5 and ${Checkpoint/.gz/.md5}"
   fi
fi

if [[ "$P4PORT" == "ssl:"* && ! -r "$P4SSLDIR/certificate.txt" ]]; then
   bail "P4PORT has SSL is enabled, but there is no certificate file [$P4SSLDIR/certificate.txt].\\nPerhaps try:\\n\\t$P4DBIN -Gc\\n\\n"
fi


[[ -x "$P4DInitScript" ]] || bail "Aborting: P4D Init script is missing or not executable: $P4DInitScript"

if [[ -r "/p4/common/config/p4_${SDPInstance}.broker.${ThisHost}.cfg" ]]; then
   P4BrokerCfg="/p4/common/config/p4_${SDPInstance}.broker.${ThisHost}.cfg"
else
   P4BrokerCfg="/p4/common/config/p4_${SDPInstance}.broker.cfg"
fi

if [[ -x "$P4BrokerInitScript" && -r "$P4BrokerCfg" ]]; then
   msg "Checking p4broker status."
   if $P4BrokerInitScript status; then
      msg "Shutting down p4broker."
      if [[ -r "$P4BrokerSystemdServiceFile" ]]; then
         sudo systemctl stop "${P4BROKERBIN##*/}" ||\
            bail "Failed to execute: sudo systemctl stop ${P4BROKERBIN##*/}"
      else
         $P4BrokerInitScript stop
      fi
   else
      msg "Verified: p4broker is down."
   fi
fi

msg "Checking p4d status."
P4DPids=$(get_pids "$P4DBIN")

if [[ -n "$P4DPids" ]]; then
   msg "Shutting down p4d."
   if [[ -r "$P4DSystemdServiceFile" ]]; then
      sudo systemctl stop "${P4DBIN##*/}" ||\
         bail "Failed to execute: sudo systemctl stop ${P4DBIN##*/}"
   else
      $P4DInitScript stop
   fi
else
   msg "Verified: No processes are running for: $P4DBIN"
fi

# Remove db.* and state* files, but not server.id or license files.
Cmd="/bin/rm -f db.* state*"
msg "Cleaning up databases with this command in $PWD:\\n$Cmd"
$Cmd || bail "Failed to clean up old databases and state files in $P4ROOT"

# Move P4LOG and P4JOURAL aside if they exist.
if [[ -r "$P4LOG" ]]; then
   Cmd="mv $P4LOG $P4LOG.moved.$(date +'%Y%m%d-%H%M')"
   msg "Moving P4LOG aside with this command:\\n$Cmd"
   $Cmd || bail "Failed to move P4LOG aside."
fi

if [[ -r "$P4JOURNAL" ]]; then
   Cmd="mv $P4JOURNAL $P4JOURNAL.moved.$(date +'%Y%m%d-%H%M')"
   msg "Moving P4JOURNAL aside with this command:\\n$Cmd"
   $Cmd || bail "Failed to move P4JOURNAL aside."
fi

# For checkpoint replay operations, determine whether case sensitivity
# flag is needed, and whether '-z' is needed.
if [[ "$Checkpoint" == *".gz" ]]; then
   CaseMode=$(zcat "$Checkpoint" | head -1 | grep -E '^@nx@ (0|2) ' | cut -d ' ' -f5)
   if [[ "$CaseMode" == "2" ]]; then
      CaseFlag='-C1'
      msg "Case-insensitive checkpoint detected."
   else
      msg "Case-sensitive checkpoint detected."
   fi

   # In P4D 2018.1, the '-z' flag should be dropped, as the need for
   # compression/decompression is determined by P4D when replying,
   # and thus '-z' should not be used (though it is supported by P4D
   # for backward compatibility, and we use it here if using an older
   # server.
   # shellcheck disable=SC2072
   [[ "$P4D_VERSION" < "2018.1" ]] && CompressFlag='-z'
else
   CaseMode=$(head -1 "$Checkpoint"| grep -E '^@nx@ (0|2) ' | cut -d ' ' -f5)
   if [[ "$CaseMode" == "2" ]]; then
      CaseFlag='-C1'
      msg "Case-insensitive checkpoint detected."
   else
      msg "Case-sensitive checkpoint detected."
   fi
fi

if [[ "${SERVERID^^}" == "P4D_EDGE_"* ]]; then
   Cmd="$P4DBIN -r $P4ROOT -K $ExcludedTables $CompressFlag $CaseFlag -jr $Checkpoint"
else
   Cmd="$P4DBIN -r $P4ROOT $CompressFlag $CaseFlag -jr $Checkpoint"
fi

msg "Replaying checkpoint to P4ROOT with this command:\\n$Cmd"
$Cmd || bail "Checkpoint replay to P4ROOT failed."

Cmd="$P4DBIN -r $P4ROOT -xu"
msg "Ensuring databases are upgraded with this command:\\n$Cmd"
$Cmd || bail "Database upgrade in P4ROOT failed."

if [[ "$LoadOfflineDB" -eq 1 ]]; then
   OfflineDB=${P4ROOT/root/offline_db}
   OfflineDBUsableFile="$OfflineDB/offline_db_usable.txt"

   cd "$OfflineDB" || bail "Could not cd to offline_db dir [$OfflineDB]."

   [[ "$PWD" == "/p4/${SDPInstance}/offline_db" ]] || bail "Unexpected OfflineDB value of: $OfflineDB"

   # Remove db.* and state* files, but not server.id or license files.
   Cmd="/bin/rm -f db.* state* $OfflineDBUsableFile"
   msg "Cleaning up databases and offline_db_usable.txt file with this command in $PWD:\\n$Cmd"
   $Cmd || bail "Failed to clean up old databases and state files in $OfflineDB"

   if [[ "${SERVERID^^}" == "P4D_EDGE_"* ]]; then
      Cmd="$P4DBIN -r $OfflineDB -K $ExcludedTables $CompressFlag $CaseFlag -jr $Checkpoint"
   else
      Cmd="$P4DBIN -r $OfflineDB $CompressFlag $CaseFlag -jr $Checkpoint"
   fi

   msg "Replaying checkpoint to offline_db with this command:\\n$Cmd"
   $Cmd || bail "Checkpoint replay to offline_db failed."

   Cmd="$P4DBIN -r $OfflineDB -xu"
   msg "Ensuring databases in offline_db are upgraded with this command:\\n$Cmd"
   $Cmd || bail "Database upgrade in offline_db failed."

   echo "Database restored successfully." > "$OfflineDBUsableFile" ||\
      bail "Failed to write this filfe: $OfflineDBUsableFile"
fi

msg "Starting p4d."
if [[ -r "$P4DSystemdServiceFile" ]]; then
   sudo systemctl start "${P4DBIN##*/}" ||\
      bail "Failed to execute: sudo systemctl start ${P4DBIN##*/}"
else
   $P4DInitScript start
fi
sleep 1

msg "Logging in."
"$P4CBIN/p4login" -v

# Now that we have loaded a checkpoint, re-load the shell environment to
# ensure that P4REPLICA is set correctly, as some db.config settings are
# determined dynamically.

# shellcheck disable=SC1091
source /p4/common/bin/p4_vars "$SDPInstance"
if [[ "$P4REPLICA" == TRUE ]]; then
   msg "Logging in service user."
   p4login -v -service

   sleep 3
   Cmd="$P4BIN -s pull -lj"
   msg "Checking replication status with this command:\\n$Cmd"
   $Cmd || bail "Replica is not replicating properly."
fi

msg "\\nAll processing completed successfully.\\n"

if [[ -x "$P4BrokerInitScript" && -r "$P4BrokerCfg" ]]; then
   if [[ "$StartBrokerWhenDone" -eq 1 ]]; then
      if [[ -r "$P4BrokerSystemdServiceFile" ]]; then
         sudo systemctl start "${P4BROKERBIN##*/}" ||\
            bail "Failed to execute: sudo systemctl start ${P4BROKERBIN##*/}"
      else
         $P4BrokerInitScript start
      fi
   else
      if [[ -r "$P4BrokerSystemdServiceFile" ]]; then
         msg "NOT starting broker. Start it manually when ready with:\\n\\tsudo systemctl start ${P4BROKERBIN##*/}\\n"
      else
         msg "NOT starting broker. Start it manually when ready with:\\n\\t$P4BrokerInitScript start\\n"
      fi
   fi
fi

msg "Checkpoint load processing took $((SECONDS/3600)) hours $((SECONDS%3600/60)) minutes $((SECONDS%60)) seconds.\\n"