Get best of the week

PostgreSQL-Rezepte: Auto-Failover und Auto-Rejoin im Docker-Schwarm

Um Auto-Failover und Auto-Rejoin im Docker-Schwarm vorzubereiten , benötigen wir Docker , Postgres , Repmgr , Pgbouncer , Runit und Gluster . Sie können auch das fertige Bild verwenden .

Zunächst organisieren wir auf zwei Hosts (besser als auf Eisen) Docker1 und Docker2 Docker-Schwarm so, dass beide Manager sind.

Installieren Sie außerdem auf beiden Hosts das verteilte Dateisystem glusterfs und hängen Sie es wie folgt auf beiden Hosts in fstab ein

localhost:/gfs /mnt/gfs glusterfs defaults,_netdev,backupvolfile-server=localhost 0 0

Erstellen Sie dann auf beiden Hosts eine Partition

docker volume create repmgr

und Netzwerk

docker network create --attachable --driver overlay docker

Sammeln Sie dann das Bild mit der Docker-Datei:

Dockerfile
FROM alpine
RUN set -ex \
    && apk add --no-cache --repository http://dl-cdn.alpinelinux.org/alpine/edge/testing --virtual .locales-rundeps \
        musl-locales \ #   (   ,      )
    && apk add --no-cache --virtual .postgresql-rundeps \
        openssh-client \ #  ssh- (   ,      )
        openssh-server \ #  ssh- (   ,    switchover)
        pgbouncer \ #  
        postgresql \ #  
        postgresql-contrib \ #     (   ,      )
        repmgr \ #  repmgr
        repmgr-daemon \ #    
        rsync \ #   (   ,      )
        runit \ #   
        shadow \ #   
        tzdata \ #   (   ,      )
    && echo done
ADD bin /usr/local/bin #  
ADD service /etc/service #    
CMD [ "runsvdir", "/etc/service" ] #   
ENTRYPOINT [ "docker_entrypoint.sh" ] #   
ENV HOME=/var/lib/postgresql
ENV GROUP=postgres \
    PGDATA="${HOME}/pg_data" \
    USER=postgres
VOLUME "${HOME}"
WORKDIR "${HOME}"
RUN set -ex \
    && sed -i -e 's|#PasswordAuthentication yes|PasswordAuthentication no|g' /etc/ssh/sshd_config \ #    
    && sed -i -e 's|#   StrictHostKeyChecking ask|   StrictHostKeyChecking no|g' /etc/ssh/ssh_config \ #    
    && echo "   UserKnownHostsFile=/dev/null" >>/etc/ssh/ssh_config \ #      
    && sed -i -e 's|postgres:!:|postgres::|g' /etc/shadow \ #    
    && chmod -R 0755 /etc/service /usr/local/bin \
    && echo done


Führen Sie nun den Dienst auf jedem Host aus

service1.sh
docker service create \
    --constraint node.hostname==docker1 \ #     
    --env GROUP_ID="$(id -g)" \ #    (       )
    --env LANG=ru_RU.UTF-8 \ #   (   ,      )
    --env TZ=Asia/Yekaterinburg \ #   (   ,      )
    --env USER_ID="$(id -u)" \ #    (       )
    --hostname tasks.repmgr1 \ #  
    --mount type=bind,source=/etc/certs,destination=/etc/certs,readonly \ #     (   ,      )
    --mount type=bind,source=/mnt/gfs/repmgr,destination=/var/lib/postgresql/gfs \ #    
    --mount type=volume,source=repmgr,destination=/var/lib/postgresql \ #  
    --name repmgr1 \ #   
    --network name=docker \ #  
    --publish target=5432,published=5432,mode=host \ #   
    --publish target=5433,published=5433,mode=host \ #   
    --replicas-max-per-node 1 \ #      
    rekgrpth/repmgr #


Führen Sie auf dem zweiten Host auch den zweiten Dienst aus. Nur anstelle von Docker1 schreiben wir Docker2 und anstelle von Repmgr1 schreiben wir Repmgr2

Wenn Docker lernen soll, den Hostnamen der Förderer aufzulösen
,
docker service create \
    --env GROUP_ID="$(id -g)" \ #    (       )
    --env LANG=ru_RU.UTF-8 \ #   (   ,      )
    --env TZ=Asia/Yekaterinburg \ #   (   ,      )
    --env USER_ID="$(id -u)" \ #    (       )
    --hostname repmgr-{{.Node.Hostname}} \ #  
    --mode global \ #       
    --mount type=bind,source=/etc/certs,destination=/etc/certs,readonly \ #     (   ,      )
    --mount type=bind,source=/mnt/gfs/repmgr,destination=/var/lib/postgresql/gfs \ #    
    --mount type=volume,source=repmgr,destination=/var/lib/postgresql \ #  
    --name repmgr \ #   
    --network name=docker \ #  
    --publish target=5432,published=5432,mode=host \ #   
    --publish target=5433,published=5433,mode=host \ #   
    rekgrpth/repmgr #


Also der Eingang zum Container

/usr/local/bin/docker_entrypoint.sh
#!/bin/sh

exec 2>&1
set -ex
if [ -n "$GROUP" ] && [ -n "$GROUP_ID" ] && [ "$GROUP_ID" != "$(id -g "$GROUP")" ]; then #          , 
    groupmod --gid "$GROUP_ID" "$GROUP" #   
    chgrp "$GROUP_ID" "$HOME" #      
fi
if [ -n "$USER" ] && [ -n "$USER_ID" ] && [ "$USER_ID" != "$(id -u "$USER")" ]; then #          , 
    usermod --uid "$USER_ID" "$USER" #   
    chown "$USER_ID" "$HOME" #      
fi
exec "$@" #


Der Supervisor startet die folgenden Dienste

1) / etc / service / ssh
, switchover.

/etc/service/ssh/run
#!/bin/sh

exec 2>&1
realpath "$0"
set -ex
ssh-keygen -A #   
exec /usr/sbin/sshd -D -e #  ssh-


, ( , )

/etc/service/ssh/log/run
#!/bin/sh

exec 2>&1
exec sed 's|^|ssh: |'



2) / etc / service / postgres


/etc/service/postgres/run
#!/bin/sh

exec 2>&1
install -d -m 0750 -o "$USER" -g "$GROUP" "$PGDATA" #    
install -d -m 1775 -o "$USER" -g "$GROUP" /run/postgresql /var/log/postgresql #     
rm -f /run/postgresql/postgres.run #   
realpath "$0"
set -ex
chmod 755 supervise
chown "$USER":"$GROUP" supervise/ok supervise/control supervise/status #    
rm -f "$PGDATA/postmaster.pid" #  pid
primary="$(test -f "$HOME/gfs/primary" && cat "$HOME/gfs/primary")" #    
test -n "$primary" || echo -n "$(hostname)" >"$HOME/gfs/primary" #    ,   -  
primary="$(test -f "$HOME/gfs/primary" && cat "$HOME/gfs/primary")" #    
test -n "$primary" # ,    
chown "$USER":"$GROUP" "$HOME/gfs/primary" #  
if [ "$primary" != "$(hostname)" ]; then #     , 
    test -d "$PGDATA/base" || /etc/service/postgres/standby #     -  
    test -f "$PGDATA/standby.signal" || /etc/service/postgres/rejoin #      ,      
else #  (  - )
    test -d "$PGDATA/base" || /etc/service/postgres/primary #     -  
fi
test -d "$PGDATA/base" # ,     
exec chpst -u "$USER":"$GROUP" -L /run/postgresql/postgres.run postmaster #


, ( , )

/etc/service/postgres/log/run
#!/bin/sh

exec 2>&1
exec sed 's|^|postgres: |'




/etc/service/postgres/standby
#!/bin/sh

exec 2>&1
realpath "$0"
set -ex
chpst -u "$USER":"$GROUP" /etc/service/repmgr/conf #    repmgr
primary="$(test -f "$HOME/gfs/primary" && cat "$HOME/gfs/primary")" #    
test -n "$primary" # ,   
chpst -u "$USER":"$GROUP" repmgr standby clone --config-file="$HOME/repmgr.conf" --verbose --fast-checkpoint --dbname="host=$primary user=repmgr dbname=repmgr connect_timeout=2" #




/etc/service/postgres/rejoin
#!/bin/sh

exec 2>&1
realpath "$0"
set -ex
primary="$(test -f "$HOME/gfs/primary" && cat "$HOME/gfs/primary")" #    
test -n "$primary" # ,   
chpst -u "$USER":"$GROUP" pg_ctl --options="-c listen_addresses=''" --wait start #    (   pg_rewind)
chpst -u "$USER":"$GROUP" pg_ctl --wait --mode=fast stop #   (   pg_rewind)
chpst -u "$USER":"$GROUP" repmgr node rejoin --config-file="$HOME/repmgr.conf" --verbose --force-rewind --no-wait --dbname="host=$primary user=repmgr dbname=repmgr connect_timeout=2" || mv -f "$PGDATA" "${PGDATA}_$(date "+%F %T")" #      (   -       )




/etc/service/postgres/primary
#!/bin/sh

exec 2>&1
realpath "$0"
set -ex
/etc/service/postgres/init #  
/etc/service/repmgr/init #  repmgr




/etc/service/postgres/init
#!/bin/sh

exec 2>&1
realpath "$0"
set -ex
cd "$PGDATA" && chpst -u "$USER":"$GROUP" initdb #  
chpst -u "$USER":"$GROUP" /etc/service/postgres/conf #   
chpst -u "$USER":"$GROUP" /etc/service/postgres/hba #




/etc/service/postgres/conf
#!/bin/sh

exec 2>&1
realpath "$0"
set -ex
cat >>"$PGDATA/postgresql.conf" <<EOF

archive_command = '/bin/true' #  
archive_mode = on #  - 
datestyle = 'iso, dmy' #   
listen_addresses = '*' #  
max_logical_replication_workers = 0 #   
max_sync_workers_per_subscription = 0 #   
ssl_ca_file = '/etc/certs/ca.pem' #   
ssl_cert_file = '/etc/certs/cert.pem' #  
ssl_key_file = '/etc/certs/key.pem' #   
ssl = on #  ssl
EOF




/etc/service/postgres/hba
#!/bin/sh

exec 2>&1
realpath "$0"
set -ex
echo >>"$PGDATA/pg_hba.conf"
ip -o -f inet addr show | awk '/scope global/ {print $4}' | sed -E 's|.\d+/|.0/|' | while read -r net; do #    
    echo "host all all $net trust" #   
done >>"$PGDATA/pg_hba.conf"



3) / etc / service / repmgr


/etc/service/repmgr/run
#!/bin/sh

exec 2>&1
install -d -m 1775 -o "$USER" -g "$GROUP" /run/postgresql #   
rm -f /run/postgresql/repmgr.run #   
test -f /run/postgresql/postgres.run || exit $? # ,    
realpath "$0"
set -ex
chmod 755 supervise
chown "$USER":"$GROUP" supervise/ok supervise/control supervise/status #    
chpst -u "$USER":"$GROUP" pg_ctl status # ,   
primary="$(test -f "$HOME/gfs/primary" && cat "$HOME/gfs/primary")" #    
test -n "$primary" # ,   
if [ "$primary" != "$(hostname)" ]; then #     , 
    test -f "$HOME/repmgr.conf" || /etc/service/repmgr/standby #    ,   
    chpst -u "$USER":"$GROUP" repmgr standby register --config-file="$HOME/repmgr.conf" --verbose --force #     
else #  (  - )
    test -f "$HOME/repmgr.conf" || /etc/service/repmgr/primary #    ,   
    test ! -f "$PGDATA/standby.signal" || /etc/service/repmgr/promote #       ,     
    chpst -u "$USER":"$GROUP" repmgr primary register --config-file="$HOME/repmgr.conf" --verbose --force #    
fi
exec chpst -u "$USER":"$GROUP" -L /run/postgresql/repmgr.run repmgrd --config-file="$HOME/repmgr.conf" --verbose --daemonize=false #   repmgr


, ( , )

/etc/service/repmgr/log/run
#!/bin/sh

exec 2>&1
exec sed 's|^|repmgr: |'


/etc/service/repmgr/standby
#!/bin/sh

exec 2>&1
realpath "$0"
set -ex
chpst -u "$USER":"$GROUP" /etc/service/repmgr/conf #




/etc/service/repmgr/primary
#!/bin/sh

exec 2>&1
realpath "$0"
set -ex
chpst -u "$USER":"$GROUP" createuser --superuser repmgr #  
chpst -u "$USER":"$GROUP" createdb repmgr --owner=repmgr #  
chpst -u "$USER":"$GROUP" /etc/service/repmgr/conf #




/etc/service/repmgr/promote
#!/bin/sh

exec 2>&1
realpath "$0"
set -ex
chpst -u "$USER":"$GROUP" repmgr standby promote --config-file="$HOME/repmgr.conf" --verbose #


/etc/service/repmgr/conf
#!/bin/sh

exec 2>&1
realpath "$0"
set -ex
cat >"$HOME/repmgr.conf" <<EOF
conninfo='host=$(hostname) user=repmgr dbname=repmgr connect_timeout=2' #      
data_directory='$(echo -n "$PGDATA")' #    
event_notification_command='/etc/service/repmgr/event "%n" "%e" "%s" "%t" "%d"' #   
failover='automatic' #   failover
failover_validation_command='/etc/service/repmgr/valid "%n" "%a"' #   
follow_command='repmgr standby follow --config-file="$(echo -n "$HOME")/repmgr.conf" --wait --upstream-node-id=%n' #   
node_id=$(hostname | grep -oE '\d+') #   
node_name='$(hostname)' #   
promote_command='repmgr standby promote --config-file="$(echo -n "$HOME")/repmgr.conf"' #   
repmgrd_service_start_command='sv start repmgr' #    
repmgrd_service_stop_command='sv force-stop repmgr' #    
service_promote_command='pg_ctl --wait promote' #    
service_reload_command='sv reload postgres' #     
service_restart_command='sv force-restart postgres' #    
service_start_command='sv start postgres' #    
service_stop_command='sv force-stop postgres' #    
ssh_options='-q -o ConnectTimeout=10' #   ssl
standby_disconnect_on_failover=true #   failover
use_replication_slots=true #   
EOF


repmgr

/etc/service/repmgr/init
#!/bin/sh

exec 2>&1
realpath "$0"
set -ex
chpst -u "$USER":"$GROUP" /etc/service/repmgr/hba #    
chpst -u "$USER":"$GROUP" cat >>"$PGDATA/postgresql.conf" <<EOF #    

hot_standby = on #   
max_replication_slots = 10 #     
max_wal_senders = 10 #    
shared_preload_libraries = 'repmgr' #  repmgr
wal_level = 'hot_standby' #    
wal_log_hints = on #    pg_rewind
EOF




/etc/service/repmgr/hba
#!/bin/sh

exec 2>&1
realpath "$0"
set -ex
echo >>"$PGDATA/pg_hba.conf"
ip -o -f inet addr show | awk '/scope global/ {print $4}' | sed -E 's|.\d+/|.0/|' | while read -r net; do #    
    echo "host replication repmgr $net trust" #    repmgr
    echo "host repmgr repmgr $net trust" #  repmgr  repmgr
done >>"$PGDATA/pg_hba.conf"
cat >>"$PGDATA/pg_hba.conf" <<EOF

local replication repmgr trust #    repmgr
host replication repmgr 127.0.0.1/32 trust #    repmgr
host replication repmgr ::1/128 trust #    repmgr

local repmgr repmgr trust #  repmgr  repmgr
host repmgr repmgr 127.0.0.1/32 trust #  repmgr  repmgr
host repmgr repmgr ::1/128 trust #  repmgr  repmgr
EOF


More articles:


All Articles