Meu servidor PostgreSQL 9.5 no Debian 8 (Jessie) continua saindo diretamente após ser iniciado via service postgresql start:

# service postgresql status
● postgresql.service - PostgreSQL RDBMS
   Loaded: loaded (/lib/systemd/system/postgresql.service; enabled)
   Active: active (exited) since Fr 2016-12-02 11:02:51 CET; 11min ago
  Process: 2360 ExecStart=/bin/true (code=exited, status=0/SUCCESS)
 Main PID: 2360 (code=exited, status=0/SUCCESS)
   CGroup: /system.slice/postgresql.service

Dez 05 16:29:24 dev systemd[1]: Starting PostgreSQL RDBMS...
Dez 05 16:29:24 dev systemd[1]: Started PostgreSQL RDBMS.

Observe o active (exited)estado. O servidor está inoperante, por exemplo, não consigo conectar-me a ele via TCP ou soquete de domínio e não consigo encontrar nenhum processo associado.

No entanto, quando inicio o PostgreSQL manualmente, ele funciona:

# sudo -u postgres /usr/lib/postgresql/9.5/bin/postgres -d 3 -D /var/lib/postgresql/9.5/main/ -c config_file=/etc/postgresql/9.5/main/postgresql.conf 

2016-12-05 16:34:32 CET [1593-1] DEBUG:  postgres: PostmasterMain: initial environment dump:
2016-12-05 16:34:32 CET [1593-2] DEBUG:  -----------------------------------------
2016-12-05 16:34:32 CET [1593-3] DEBUG:     LC_PAPER=de_DE.UTF-8
2016-12-05 16:34:32 CET [1593-4] DEBUG:     LC_ADDRESS=de_DE.UTF-8
2016-12-05 16:34:32 CET [1593-5] DEBUG:     LC_MONETARY=C
2016-12-05 16:34:32 CET [1593-6] DEBUG:     TERM=xterm-256color
2016-12-05 16:34:32 CET [1593-7] DEBUG:     LC_NUMERIC=C
2016-12-05 16:34:32 CET [1593-8] DEBUG:     LC_TELEPHONE=de_DE.UTF-8
2016-12-05 16:34:32 CET [1593-9] DEBUG:     LS_COLORS=rs=0:di=01;34:ln=01;36:mh=00:pi=40;33:so=01;35:do=01;35:bd=40;33;01:cd=40;33;01:or=40;31;01:su=37;41:sg=30;43:ca=30;41:tw=30;42:ow=34;42:st=37;44:ex=01;32:*.tar=01;31:*.tgz=01;31:*.arc=01;31:*.arj=01;31:*.taz=01;31:*.lha=01;31:*.lz4=01;31:*.lzh=01;31:*.lzma=01;31:*.tlz=01;31:*.txz=01;31:*.tzo=01;31:*.t7z=01;31:*.zip=01;31:*.z=01;31:*.Z=01;31:*.dz=01;31:*.gz=01;31:*.lrz=01;31:*.lz=01;31:*.lzo=01;31:*.xz=01;31:*.bz2=01;31:*.bz=01;31:*.tbz=01;31:*.tbz2=01;31:*.tz=01;31:*.deb=01;31:*.rpm=01;31:*.jar=01;31:*.war=01;31:*.ear=01;31:*.sar=01;31:*.rar=01;31:*.alz=01;31:*.ace=01;31:*.zoo=01;31:*.cpio=01;31:*.7z=01;31:*.rz=01;31:*.cab=01;31:*.jpg=01;35:*.jpeg=01;35:*.gif=01;35:*.bmp=01;35:*.pbm=01;35:*.pgm=01;35:*.ppm=01;35:*.tga=01;35:*.xbm=01;35:*.xpm=01;35:*.tif=01;35:*.tiff=01;35:*.png=01;35:*.svg=01;35:*.svgz=01;35:*.mng=01;35:*.pcx=01;35:*.mov=01;35:*.mpg=01;35:*.mpeg=01;35:*.m2v=01;35:*.mkv=01;35:*.webm=01;35:*.ogm=01;35:*.mp4=01;35:*.m4v=01;35:*.mp4v=01;35:*.vob=01;35:*.qt=01;35:*.nuv=01;35:*.wmv=01;35:*.asf=01;35:*.rm=01;35:*.rmvb=01;35:*.flc=01;35:*.avi=01;35:*.fli=01;35:*.flv=01;35:*.gl=01;35:*.dl=01;35:*.xcf=01;35:*.xwd=01;35:*.yuv=01;35:*.cgm=01;35:*.emf=01;35:*.axv=01;35:*.anx=01;35:*.ogv=01;35:*.ogx=01;35:*.aac=00;36:*.au=00;36:*.flac=00;36:*.m4a=00;36:*.mid=00;36:*.midi=00;36:*.mka=00;36:*.mp3=00;36:*.mpc=00;36:*.ogg=00;36:*.ra=00;36:*.wav=00;36:*.axa=00;36:*.oga=00;36:*.spx=00;36:*.xspf=00;36:
2016-12-05 16:34:32 CET [1593-10] DEBUG:    PATH=/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin
2016-12-05 16:34:32 CET [1593-11] DEBUG:    LC_IDENTIFICATION=de_DE.UTF-8
2016-12-05 16:34:32 CET [1593-12] DEBUG:    LANG=en_US.UTF-8
2016-12-05 16:34:32 CET [1593-13] DEBUG:    LC_MEASUREMENT=de_DE.UTF-8
2016-12-05 16:34:32 CET [1593-14] DEBUG:    LC_TIME=C
2016-12-05 16:34:32 CET [1593-15] DEBUG:    LC_NAME=de_DE.UTF-8
2016-12-05 16:34:32 CET [1593-16] DEBUG:    SHELL=/bin/bash
2016-12-05 16:34:32 CET [1593-17] DEBUG:    MAIL=/var/mail/postgres
2016-12-05 16:34:32 CET [1593-18] DEBUG:    LOGNAME=postgres
2016-12-05 16:34:32 CET [1593-19] DEBUG:    USER=postgres
2016-12-05 16:34:32 CET [1593-20] DEBUG:    USERNAME=postgres
2016-12-05 16:34:32 CET [1593-21] DEBUG:    HOME=/var/lib/postgresql
2016-12-05 16:34:32 CET [1593-22] DEBUG:    SUDO_COMMAND=/usr/lib/postgresql/9.5/bin/postgres -d 3 -D /var/lib/postgresql/9.5/main/ -c config_file=/etc/postgresql/9.5/main/postgresql.conf
2016-12-05 16:34:32 CET [1593-23] DEBUG:    SUDO_USER=root
2016-12-05 16:34:32 CET [1593-24] DEBUG:    SUDO_UID=0
2016-12-05 16:34:32 CET [1593-25] DEBUG:    SUDO_GID=0
2016-12-05 16:34:32 CET [1593-26] DEBUG:    PGLOCALEDIR=/usr/share/locale
2016-12-05 16:34:32 CET [1593-27] DEBUG:    PGSYSCONFDIR=/etc/postgresql-common
2016-12-05 16:34:32 CET [1593-28] DEBUG:    LC_COLLATE=en_US.UTF-8
2016-12-05 16:34:32 CET [1593-29] DEBUG:    LC_CTYPE=en_US.UTF-8
2016-12-05 16:34:32 CET [1593-30] DEBUG:    LC_MESSAGES=en_US.UTF-8
2016-12-05 16:34:32 CET [1593-31] DEBUG:  -----------------------------------------
2016-12-05 16:34:32 CET [1593-32] DEBUG:  invoking IpcMemoryCreate(size=23887872)
2016-12-05 16:34:32 CET [1593-33] DEBUG:  mmap(25165824) with MAP_HUGETLB failed, huge pages disabled: Cannot allocate memory
2016-12-05 16:34:32 CET [1593-34] DEBUG:  SlruScanDirectory invoking callback on pg_notify/0000
2016-12-05 16:34:32 CET [1593-35] DEBUG:  removing file "pg_notify/0000"
2016-12-05 16:34:32 CET [1593-36] DEBUG:  dynamic shared memory system will support 288 segments
2016-12-05 16:34:32 CET [1593-37] DEBUG:  created dynamic shared memory control segment 1173813643 (2316 bytes)
2016-12-05 16:34:32 CET [1593-38] DEBUG:  max_safe_fds = 983, usable_fds = 1000, already_open = 7
2016-12-05 16:34:32 CET [1593-39] LOG:  redirecting log output to logging collector process
2016-12-05 16:34:32 CET [1593-40] HINT:  Future log output will appear in directory "/var/log/postgresql".

(server is now up and running and can be connected to)

Os logs /var/log/postgresqlnão contêm nada suspeito. Aqui está a saída para uma execução manual de trabalho:

2016-12-05 16:48:07 CET [1700-1] LOG:  database system was shut down at 2016-12-05 16:36:04 CET
2016-12-05 16:48:07 CET [1700-2] DEBUG:  checkpoint record is at 0/13823548
2016-12-05 16:48:07 CET [1700-3] DEBUG:  redo record is at 0/13823548; shutdown TRUE
2016-12-05 16:48:07 CET [1700-4] DEBUG:  next transaction ID: 0/14019; next OID: 35693
2016-12-05 16:48:07 CET [1700-5] DEBUG:  next MultiXactId: 1; next MultiXactOffset: 0
2016-12-05 16:48:07 CET [1700-6] DEBUG:  oldest unfrozen transaction ID: 617, in database 1
2016-12-05 16:48:07 CET [1700-7] DEBUG:  oldest MultiXactId: 1, in database 1
2016-12-05 16:48:07 CET [1700-8] DEBUG:  commit timestamp Xid oldest/newest: 0/0
2016-12-05 16:48:07 CET [1700-9] DEBUG:  transaction ID wrap limit is 2147484264, limited by database with OID 1
2016-12-05 16:48:07 CET [1700-10] DEBUG:  MultiXactId wrap limit is 2147483648, limited by database with OID 1
2016-12-05 16:48:07 CET [1700-11] DEBUG:  starting up replication slots
2016-12-05 16:48:07 CET [1700-12] DEBUG:  MultiXactId wrap limit is 2147483648, limited by database with OID 1
2016-12-05 16:48:07 CET [1700-13] LOG:  MultiXact member wraparound protections are now enabled
2016-12-05 16:48:07 CET [1700-14] DEBUG:  MultiXact member stop limit is now 4294914944 based on MultiXact 1
2016-12-05 16:48:07 CET [1700-15] DEBUG:  shmem_exit(0): 1 before_shmem_exit callbacks to make
2016-12-05 16:48:07 CET [1700-16] DEBUG:  shmem_exit(0): 3 on_shmem_exit callbacks to make
2016-12-05 16:48:07 CET [1700-17] DEBUG:  proc_exit(0): 2 callbacks to make
2016-12-05 16:48:07 CET [1700-18] DEBUG:  exit(0)
2016-12-05 16:48:07 CET [1700-19] DEBUG:  shmem_exit(-1): 0 before_shmem_exit callbacks to make
2016-12-05 16:48:07 CET [1700-20] DEBUG:  shmem_exit(-1): 0 on_shmem_exit callbacks to make
2016-12-05 16:48:07 CET [1700-21] DEBUG:  proc_exit(-1): 0 callbacks to make
2016-12-05 16:48:07 CET [1698-41] LOG:  database system is ready to accept connections
2016-12-05 16:48:07 CET [1705-1] DEBUG:  removing permanent stats file "pg_stat/db_12381.stat"
2016-12-05 16:48:07 CET [1704-1] LOG:  autovacuum launcher started
2016-12-05 16:48:07 CET [1704-2] DEBUG:  InitPostgres
2016-12-05 16:48:07 CET [1705-2] DEBUG:  removing permanent stats file "pg_stat/db_0.stat"
2016-12-05 16:48:07 CET [1704-3] DEBUG:  StartTransaction
2016-12-05 16:48:07 CET [1705-3] DEBUG:  removing permanent stats file "pg_stat/global.stat"
2016-12-05 16:48:07 CET [1704-4] DEBUG:  name: unnamed; blockState:       DEFAULT; state: INPROGR, xid/subid/cid: 0/1/0, nestlvl: 1, children: 
2016-12-05 16:48:07 CET [1704-5] DEBUG:  CommitTransaction
2016-12-05 16:48:07 CET [1704-6] DEBUG:  name: unnamed; blockState:       STARTED; state: INPROGR, xid/subid/cid: 0/1/0, nestlvl: 1, children: 
2016-12-05 16:48:07 CET [1705-4] DEBUG:  received inquiry for database 0
2016-12-05 16:48:07 CET [1705-5] DEBUG:  writing stats file "/var/run/postgresql/9.5-main.pg_stat_tmp/global.stat"
2016-12-05 16:48:07 CET [1705-6] DEBUG:  writing stats file "/var/run/postgresql/9.5-main.pg_stat_tmp/db_0.stat"
2016-12-05 16:48:07 CET [1701-1] DEBUG:  checkpointer updated shared memory configuration values

No caso de uma execução sem êxito via service postgresql start, nenhum registro é gravado /var/log/postgresql. Da mesma forma, não consigo encontrar nenhuma mensagem relacionada ao PostgreSQL /var/log/messages.

Instalei o PostgreSQL através do repositório oficial:

# apt-cache policy postgresql-9.5
postgresql-9.5:
  Installed: 9.5.5-1.pgdg80+1
  Candidate: 9.5.5-1.pgdg80+1
  Version table:
 *** 9.5.5-1.pgdg80+1 0
        500 http://apt.postgresql.org/pub/repos/apt/ jessie-pgdg/main amd64 Packages
        100 /var/lib/dpkg/status

Tudo isso está em uma VM do VirtualBox.

Atualização 1

A unidade systemd parece pertencer ao postgresql-commonpacote:

# find /lib/systemd -iname "*postgres*"
/lib/systemd/system/postgresql.service
/lib/systemd/system/postgresql@.service
/lib/systemd/system-generators/postgresql-generator

# dpkg -S postgresql.service
postgresql-common: /lib/systemd/system/postgresql.service

# dpkg -S postgresql@.service
postgresql-common: /lib/systemd/system/postgresql@.service

# dpkg -S postgresql-generator
postgresql-common: /lib/systemd/system-generators/postgresql-generator

Aqui está o conteúdo deles:

/lib/systemd/system/postgresql.service

# systemd service for managing all PostgreSQL clusters on the system. This
# service is actually a systemd target, but we are using a service since
# targets cannot be reloaded.

[Unit]
Description=PostgreSQL RDBMS

[Service]
Type=oneshot
ExecStart=/bin/true
ExecReload=/bin/true
RemainAfterExit=on

[Install]
WantedBy=multi-user.target

/lib/systemd/system/postgresql@.service

# systemd service template for PostgreSQL clusters. The actual instances will
# be called "postgresql@version-cluster", e.g. "postgresql@9.3-main". The
# variable %i expands to "version-cluster", %I expands to "version/cluster".
# (%I breaks for cluster names containing dashes.)

[Unit]
Description=PostgreSQL Cluster %i
ConditionPathExists=/etc/postgresql/%I/postgresql.conf
PartOf=postgresql.service
ReloadPropagatedFrom=postgresql.service
Before=postgresql.service

[Service]
Type=forking
# @: use "postgresql@%i" as process name
ExecStart=@/usr/bin/pg_ctlcluster postgresql@%i --skip-systemctl-redirect %i start
ExecStop=/usr/bin/pg_ctlcluster --skip-systemctl-redirect -m fast %i stop
ExecReload=/usr/bin/pg_ctlcluster --skip-systemctl-redirect %i reload
PIDFile=/var/run/postgresql/%i.pid
SyslogIdentifier=postgresql@%i
# prevent OOM killer from choosing the postmaster (individual backends will
# reset the score to 0)
OOMScoreAdjust=-900
# restarting automatically will prevent "pg_ctlcluster ... stop" from working,
# so we disable it here. Also, the postmaster will restart by itself on most
# problems anyway, so it is questionable if one wants to enable external
# automatic restarts.
#Restart=on-failure
# (This should make pg_ctlcluster stop work, but doesn't:)
#RestartPreventExitStatus=SIGINT SIGTERM

[Install]
WantedBy=multi-user.target

/lib/systemd/system-generators/postgresql-generator

#!/bin/sh

# This systemd generator creates dependency symlinks that make all PostgreSQL
# clusters with "auto" in their start.conf file be started/stopped/reloaded
# when postgresql.service is started/stopped/reloaded.

set -eu

gendir="$1"
wantdir="$1/postgresql.service.wants"
pgservice="/lib/systemd/system/postgresql@.service"

mkdir -p "$wantdir"

for conf in /etc/postgresql/*/*/postgresql.conf; do
    test -e "$conf" || continue
    dir="${conf%/*}"

    # evaluate start.conf
    if [ -e "$dir/start.conf" ]; then
        start=$(sed 's/#.*$//; /^[[:space:]]*$/d; s/^\s*//; s/\s*$//' "$dir/start.conf")
    else
        start=auto
    fi
    [ "$start" = "auto" ] || continue

    verdir="${dir%/*}"
    version="${verdir##*/}"
    cluster="${dir##*/}"
    ln -s "$pgservice" "$wantdir/postgresql@$version-$cluster.service"
done

exit 0

postgresql-commonfoi instalado a partir do repositório oficial do PostgreSQL como uma dependência do postgresql-9.5pacote:

# apt-cache policy postgresql-common
postgresql-common:
  Installed: 177.pgdg80+1
  Candidate: 177.pgdg80+1
  Version table:
 *** 177.pgdg80+1 0
        500 http://apt.postgresql.org/pub/repos/apt/ jessie-pgdg/main amd64 Packages
        100 /var/lib/dpkg/status
     165+deb8u1 0
        500 http://httpredir.debian.org/debian/ jessie/main amd64 Packages

Atualização 2

Conforme solicitado, aqui está a saída do strace -f service postgresql start Pastebin (devido ao seu tamanho).

Há algumas coisas acontecendo aqui. Se você chamar um serviço systemd com um @, é um sinal para iniciar uma instância em vez do serviço principal. Colocar @9.5-mainestá informando os parâmetros da instância a serem usados.

O arquivo de serviço do postgres que você está vendo nada mais é do que um espaço reservado. Há um diretório no sistema (normalmente em algum lugar /run) chamado postgresql.wants.dque o systemd procurará quando postgresql.servicefor iniciado.

Qualquer coisa nesse diretório será iniciada juntamente com postgresql.servicea informação desejada.

Esse diretório é preenchido por meio de um arquivo gerador de script de shell, pois tudo /runé um tmpfs. Esse é o script de shell que você citou acima.

Se eu tivesse que arriscar um palpite sobre o que está acontecendo lá, através da lógica do script de shell, você provavelmente terá o arquivo /etc/postgresql/9.5/main/start.confdefinido manual.

Isso significa que o gerador o ignora e postgresql.wants.dnunca é adicionado com esse nome de instância.

De acordo com a lógica do script, simplesmente removê-lo tem o efeito certo de torná-lo automático.

Você também pode configurá-lo como autoe fazer a daemon-reloadpara regenerar a configuração necessária.

Em 26 de maio de 2017, o problema ainda existe no Debian 8 (tanto quanto eu posso ver): O PostgreSQL falha ao iniciar na inicialização.

A solução, por enquanto, é emitir:

systemctl daemon-reload
systemctl restart postgresql

manualmente na inicialização e tudo deve ficar bem. Agradecemos a Florian Brucker por apontar isso acima.

Aparentemente, o bug está no systemd e, esperançosamente, deve ser corrigido em breve.