automate/020_sanoid.sh

514 lines
13 KiB
Bash
Executable File

#!/usr/bin/env bash
# set -x
sudo apt install -y \
sanoid
# git \
# dh-make \
# libconfig-inifiles-perl \
echo "sanoid provides: sanoid, syncoid, findoid in /usr/sbin"
echo "defaults are in /usr/share/sanoid/sanoid.defaults.conf"
echo "override them in /etc/sanoid/sanoid.conf"
echo "Installed files:"
cat <<-EOF
syncoid /usr/sbin/
sanoid /usr/sbin/
findoid /usr/sbin/
sanoid.defaults.conf /usr/share/sanoid/
debian/sanoid-prune.service /lib/systemd/system
CHANGELIST /usr/share/doc/sanoid/changelog
EOF
# Both sanoid and syncoid are oneshot processes so it makes little sense to
# provide an init file, cron is just fine. In this case the systemd file is there
# because systemd decided to manage cron tasks.
#
# Cronjob for non-systemd systems: every 15 minutes.
# If you require a different interval, you will need to disable the
# timer or the cron job according to your system configuration.
conf_print_sanoid_cron() {
cat <<-EOF
PATH=/usr/local/sbin:/usr/local/bin:/sbin:/bin:/usr/sbin:/usr/bin
*/15 * * * * root [ -f /etc/sanoid/sanoid.conf ] && if [ ! -d /run/systemd/system ]; then TZ=UTC /usr/sbin/sanoid --cron --quiet; fi
EOF
}
conf_print_sanoid_cron | sudo tee /etc/cron.d/sanoid
# Do the same for syncoid for backups.
conf_print_syncoid_cron() {
cat <<-EOF
PATH=/usr/local/sbin:/usr/local/bin:/sbin:/bin:/usr/sbin:/usr/bin
*/15 * * * * root [ -f /etc/sanoid/syncoid.conf ] && if [ ! -d /run/systemd/system ]; then TZ=UTC /usr/sbin/syncoid --cron --quiet; fi
EOF
}
# conf_print_syncoid_cron | sudo tee /etc/cron.d/syncoid
# Copied from /usr/share/doc/sanoid/example.conf
sudo mkdir -p /etc/sanoid/
conf_print_sanoid() {
cat <<-EOF
#-- Organisational datasets
# DATA
[rpool/DATA]
use_template = ignore
recursive = yes
process_children_only = yes
# ROOT
[rpool/ROOT]
use_template = ignore
recursive = yes
process_children_only = yes
#-- END
# These datasets contain the docker zsh backing store "graph" (layers).
# Just restore broken container with docker-compose down && docker-compose up
[rpool/docker]
use_template = ignore
recursive = yes
# Docker persistent data
[rpool/DATA/docker-volumes]
use_template = docker-persistent
recursive = yes
[rpool/archive]
use_template = ignore
recursive = yes
[rpool/swap]
use_template = ignore
recursive = yes
[rpool/tmp]
use_template = ignore
recursive = yes
[rpool/ROOT/devuan-1]
use_template = root
recursive = yes
# rpool/ROOT/devuan-1/opt
# rpool/ROOT/devuan-1/usr
# rpool/ROOT/devuan-1/usr_local
# rpool/ROOT/devuan-1/usr_share
# rpool/ROOT/devuan-1/var
# rpool/ROOT/devuan-1/var_lib
# Specific override for Virtual Machines to use scripts
[rpool/ROOT/devuan-1/var_lib_virt]
use_template = root
recursive = yes
pre_snapshot_script = /usr/local/bin/sanoid_virt-freeze-all.sh
post_snapshot_script = /usr/local/bin/sanoid_virt-thaw-all.sh
# -- User Data --
[rpool/home]
use_template = production
recursive = yes
[rpool/space]
use_template = production
#############################
# templates below this line #
#############################
[template_production]
frequently = 0
hourly = 36
daily = 30
monthly = 3
yearly = 0
autosnap = yes
autoprune = yes
[template_root]
# Root changes fast; shorter history often suffices
hourly = 24
daily = 7
monthly = 1
yearly = 0
autosnap = yes
autoprune = yes
[template_ignore]
autoprune = no
autosnap = no
monitor = no
##############################
# Docker Persistent Template #
##############################
[template_docker-persistent]
# Frequent snapshots for active databases/configs
frequently = 0
hourly = 24
daily = 7
monthly = 1
yearly = 0
# Safety checks
autosnap = yes
autoprune = yes
# Don't take a snapshot if the dataset hasn't changed
# (Saves metadata overhead)
# skip_hourless = yes
pre_snapshot_script = /usr/local/bin/sanoid_zfs-skip-empty.sh
#-- END
[template_backup]
autoprune = yes
frequently = 0
hourly = 30
daily = 90
monthly = 12
yearly = 0
### don't take new snapshots - snapshots on backup
### datasets are replicated in from source, not
### generated locally
autosnap = no
### monitor hourlies and dailies, but don't warn or
### crit until they're over 48h old, since replication
### is typically daily only
hourly_warn = 2880
hourly_crit = 3600
daily_warn = 48
daily_crit = 60
#-- END
[template_hotspare]
autoprune = yes
frequently = 0
hourly = 30
daily = 90
weekly = 4
monthly = 3
yearly = 0
### don't take new snapshots - snapshots on backup
### datasets are replicated in from source, not
### generated locally
autosnap = no
### monitor hourlies and dailies, but don't warn or
### crit until they're over 4h old, since replication
### is typically hourly only
hourly_warn = 4h
hourly_crit = 6h
daily_warn = 2d
daily_crit = 4d
EOF
}
conf_print_sanoid | sudo tee /etc/sanoid/sanoid.conf
########################
# Pre-snapshot scripts #
########################
# In ZFS, even if no data has changed, creating a snapshot still consumes a
# small amount of space for metadata and adds an entry to the ZFS history.
# If you have hundreds of datasets being snapshotted every 15 minutes, this
# "metadata bloat" can make commands like zfs list -t snapshot feel sluggish
# over time. If you think this is an issue for ypu use zfs-skip-empty.sh as
# a pre_snapshot_script
conf_print_skip_empty() {
cat <<'EOF'
#!/bin/bash
# Usage: ./sanoid-threshold.sh <dataset> <threshold_size>
# or
# Add this to you /etc/sanoid.conf to fire this script.
# [tank/important_data]
# use_template = production
# # Only snapshot if more than 5MB changed
# pre_snapshot_script = /usr/local/bin/sanoid-threshold.sh 5M
DATASET=$1
RAW_THRESHOLD=$2
convert_to_bytes() {
local number=$(echo "$1" | grep -oE '^[0-9.]+')
local unit=$(echo "$1" | grep -oI '[KMGPT]' | tr '[:lower:]' '[:upper:]')
case "$unit" in
K) awk "BEGIN { printf \"%.0f\", $number * 1024 }" ;;
M) awk "BEGIN { printf \"%.0f\", $number * 1024^2 }" ;;
G) awk "BEGIN { printf \"%.0f\", $number * 1024^3 }" ;;
T) awk "BEGIN { printf \"%.0f\", $number * 1024^4 }" ;;
*) printf "%.0f" "$number" ;;
esac
}
if [[ -z "$DATASET" || -z "$RAW_THRESHOLD" ]]; then
logger -t sanoid "Threshold Error: Usage: $0 <dataset> <threshold>"
exit 1
fi
if ! zfs list -H "$DATASET" >/dev/null 2>&1; then
logger -t sanoid "Threshold Error: Dataset $DATASET not found."
exit 1
fi
THRESHOLD=$(convert_to_bytes "$RAW_THRESHOLD")
WRITTEN_BYTES=$(zfs get -Hp -o value written "$DATASET")
if [[ "$WRITTEN_BYTES" -lt "$THRESHOLD" ]]; then
WRITTEN_HUMAN=$(zfs get -H -o value written "$DATASET")
# Optional: Comment out the logger below if your logs get too noisy
logger -t sanoid "Skipping $DATASET: Written $WRITTEN_HUMAN < Threshold $RAW_THRESHOLD."
exit 1
fi
exit 0
EOF
}
conf_print_skip_hourless | sudo tee /usr/local/bin/sanoid_zfs-skip-empty.sh
chmod +x /usr/local/bin/sanoid_zfs-skip-empty.sh
# VM Consistency (The "Freeze/Thaw" Logic)
# The inclusion of virt-freeze-all.sh and virt-thaw-all.sh to ensure data integrity:
# * Pre-snapshot: virsh domfsfreeze tells the Guest OS (via qemu-guest-agent) to flush its write buffers and temporarily pause I/O.
# * Snapshot: Sanoid takes an atomic ZFS snapshot.
# * Post-snapshot: virsh domfsthaw resumes I/O.
# Result: You get an "application-consistent" backup rather than a "crash-consistent" one.
# Give sudo access to virsh or is part of the libvirt group.
# qemu-guest-agent must be running in the vm
conf_print_virt_freeze() {
cat <<'EOF'
#!/bin/sh
# /usr/local/bin/virt-freeze.sh
VM_NAME="$1"
if [ -z "$VM_NAME" ]; then
echo "Usage: $0 <domain_name>"
exit 1
fi
# Check if the VM is running
STATE=$(/usr/bin/virsh domstate "$VM_NAME" 2>/dev/null)
if [ "$STATE" != "running" ]; then
echo "VM $VM_NAME is not running or does not exist. Skipping freeze."
exit 0
fi
echo "Freezing filesystems for $VM_NAME..."
# domfsfreeze returns the number of frozen filesystems on success
if /usr/bin/virsh domfsfreeze "$VM_NAME" > /dev/null; then
echo "Successfully frozen $VM_NAME."
else
echo "Error: Failed to freeze $VM_NAME. Ensure qemu-guest-agent is active."
exit 1
fi
EOF
}
conf_print_virt_freeze | sudo tee /usr/local/bin/sanoid_virt-freeze.sh
sudo chmod +x /usr/local/bin/sanoid_virt-freeze.sh
#---
conf_print_virt_unfreeze() {
cat <<'EOF'
#!/bin/sh
# /usr/local/bin/virt-unfreeze.sh
VM_NAME="$1"
if [ -z "$VM_NAME" ]; then
echo "Usage: $0 <domain_name>"
exit 1
fi
# Check if the VM is running
STATE=$(/usr/bin/virsh domstate "$VM_NAME" 2>/dev/null)
if [ "$STATE" != "running" ]; then
echo "VM $VM_NAME is not running. Skipping unfreeze."
exit 0
fi
echo "Thawing filesystems for $VM_NAME..."
# domfsthaw returns the number of thawed filesystems on success
if /usr/bin/virsh domfsthaw "$VM_NAME" > /dev/null; then
echo "Successfully thawed $VM_NAME."
else
echo "Error: Failed to thaw $VM_NAME."
exit 1
fi
EOF
}
conf_print_virt_unfreeze | sudo tee /usr/local/bin/sanoid_virt-unfreeze.sh
sudo chmod +x /usr/local/bin/sanoid_virt-unfreeze.sh
#---
conf_print_virt_thaw_all() {
cat <<'EOF'
#!/bin/bash
# /usr/local/bin/virt-thaw-all.sh
# 1. Get running VM names, filtering out empty lines with awk
RUNNING_VMS=$(/usr/bin/virsh list --state-running --name | awk 'NF')
if [ -z "$RUNNING_VMS" ]; then
echo "No running VMs found."
exit 0
fi
# 2. Loop through each running VM
for VM_NAME in $RUNNING_VMS; do
echo "Processing $VM_NAME..."
# Use the native thaw command.
# It handles the guest agent communication for you.
if /usr/bin/virsh domfsthaw "$VM_NAME" > /dev/null 2>&1; then
echo "Successfully thawed $VM_NAME."
else
# If native fails, we capture the error for the user
ERROR=$(/usr/bin/virsh domfsthaw "$VM_NAME" 2>&1)
echo "Error thawing $VM_NAME: $ERROR"
fi
done
echo "Finished processing all VMs."
EOF
}
conf_print_virt_thaw_all | sudo tee /usr/local/bin/sanoid_virt-thaw-all.sh
sudo chmod +x /usr/local/bin/sanoid_virt-thaw-all.sh
#---
conf_print_virt_freeze-all() {
cat <<'EOF'
#!/bin/bash
# 1. Get running VM names, filtering out empty lines
RUNNING_VMS=$(/usr/bin/virsh list --state-running --name | awk 'NF')
if [ -z "$RUNNING_VMS" ]; then
echo "No running VMs found."
exit 0
fi
# 2. Loop through each running VM
for VM_NAME in $RUNNING_VMS; do
echo "Attempting to freeze $VM_NAME..."
# Using the native virsh command is generally cleaner
# It returns the number of frozen filesystems on success
if /usr/bin/virsh domfsfreeze "$VM_NAME" > /dev/null 2>&1; then
echo "Successfully frozen $VM_NAME."
else
echo "Error: Could not freeze $VM_NAME. Check if QEMU Guest Agent is running."
fi
done
echo "Finished processing all VMs."
EOF
}
conf_print_virt_freeze-all | sudo tee /usr/local/bin/sanoid_virt-freeze-all.sh
sudo chmod +x /usr/local/bin/sanoid_virt-freeze-all.sh
#---
# exit 0
#
# # build a debian package from the git repo
# USER="jimsalterjrs"
# PROJECT="sanoid"
# LICENSE="mit"
# SECTION="admin"
# HOMEPAGE="https://openoid.net"
# BUILD_HOME="/var/tmp/sanoid-git_build"
# #VERSION="2.10.0"
# PACKAGE="sanoid"
# VERSION=$(curl "https://api.github.com/repos/${USER}/${PROJECT}/tags?per_page=5" | jq -r '.[0] | .name')
# VERSION=${VERSION:1} # strip the preceding 'v'
# DEBEMAIL="person@company.tld"
# DEBFULLNAME="Testy McTester"
# DESCRIPTION="Policy-driven snapshot management tool for ZFS filesystems."
# LONG_DESCRIPTION="Create, thin, and monitor snapshots and pool health from a single configuration file."
#
# RSS_FEED="https://github.com/${USER}/${PROJECT}/releases.atom"
# # xdg-open ${RSS_FEED}
# echo ${RSS_FEED}
#
# # Make the BUILD_HOME directory and clone upstream
# mkdir -p ${BUILD_HOME}
# cd ${BUILD_HOME} || exit
#
# git clone http://github.com/${USER}/${PROJECT}.git ${PACKAGE}-${VERSION}
#
# cd ${PACKAGE}-${VERSION} || exit
#
# mkdir -p m4
#
# cat <<EOF >.git/gbp.conf
# [DEFAULT]
# # this is the upstream-branch:
# upstream-branch=master
# EOF
#
# # debianize the BUILD_HOME directory
# echo ${DEBFULLNAME}
# dh_make \
# --createorig \
# --single \
# --native \
# --copyright ${LICENSE} \
# --yes
#
# cp README.rst debian/README
# rm debian/README.source
#
# # Customize the debian directory values
# #sed -i 's,^\(Description: \).*,\1'${DESCRIPTION}',' debian/control
# sed -i "s,^\(Description: \).*,\1${DESCRIPTION}," debian/control
# sed -i 's,^\(Section: \).*,\1'${SECTION}',' debian/control
# sed -i "s,^\(Maintainer: \).*,\1'${DEBFULLNAME}\ \<${DEBEMAIL}\>'," debian/control
# sed -i 's,^\(Homepage: \).*,\1'${HOMEPAGE}',' debian/control
# sed -i "s/.*insert\ long.*/${LONG_DESCRIPTION}/" debian/control
#
# echo ****************
# cat debian/control
# echo ****************
#
# BUILD_OPTIONS=""
#
# cat <<EOF >debian/rules
# #!/usr/bin/make -f
# %:
# dh \$@
#
# override_dh_auto_configure:
# dh_auto_configure -- ${BUILD_OPTIONS}
# EOF
#
# cat <<'EOF' >debian/watch
# version=4
# opts=filenamemangle=s/.+\/v?(\d\S+)\.tar\.gz/${PROJECT}-$1\.tar\.gz/ \
# http://github.com/${USER}/${PROJECT}/tags .*/v?(\d\S+)\.tar\.gz
# EOF
#
# cp -a packages/debian/* debian
#
# dpkg-buildpackage -us -uc -b
#
# echo ${PACKAGE}_${VERSION}_amd64.deb
#
# sudo dpkg -i ../${PACKAGE}_${VERSION}_amd64.deb