Trouble performing backups

[inful]

I get the following error when trying to do a backup:

pg_basebackup: could not connect to server: FATAL:  number of requested standby connections exceeds max_wal_senders (currently 0)

[inful]

Got it working, but had to set the following manually in postgresql.conf (in persistent data dir):

wal_level = hot_standby
max_wal_senders = 16
wal_keep_segments = 4
max_replication_slots = 2

I was under the impression that it should work out of the box:)

[sameersbn]

Yes, it should have worked out of the box as long as a replication user/pass exists on the primary instance.

Can you share the image version that you are/were using. Are you able to reproduce the issue with a fresh setup. That you help me track down the issue.

(Sorry for the late reply.)

[inful]

I am running 9.4-22 (incrementally updated from 9.4). I can reproduce this by running the following compose file:

version: '2'
services:
  master:
    image: sameersbn/postgresql:9.4
    environment:
      REPLICATION_USER: 'repl'
      REPLICATION_PASS: 'repl'
#  slave:
#    image: sameersbn/postgresql:9.4-22
#    environment:
#      REPLICATION_USER: 'repl'
#      REPLICATION_PASS: 'repl'
#      REPLICATION_MODE: 'slave'
#      REPLICATION_HOST: 'master'

Stopping it, and running it again with:

version: '2'
services:
  master:
    image: sameersbn/postgresql:9.4-22
    environment:
      REPLICATION_USER: 'repl'
      REPLICATION_PASS: 'repl'
  slave:
    image: sameersbn/postgresql:9.4-22
    environment:
      REPLICATION_USER: 'repl'
      REPLICATION_PASS: 'repl'
      REPLICATION_MODE: 'slave'
      REPLICATION_HOST: 'master'

This will result in the following output:

Recreating pgtest_master_1
Creating pgtest_slave_1
Attaching to pgtest_slave_1, pgtest_master_1
slave_1   | Initializing datadir...
master_1  | Initializing datadir...
slave_1   | Initializing certdir...
slave_1   | Initializing logdir...
slave_1   | Initializing rundir...
slave_1   | Setting resolv.conf ACLs...
master_1  | Initializing certdir...
master_1  | Initializing logdir...
master_1  | Initializing rundir...
master_1  | Setting resolv.conf ACLs...
master_1  | ‣ Setting postgresql.conf parameter: data_directory = '/var/lib/postgresql/9.4/main'
master_1  | ‣ Setting postgresql.conf parameter: log_directory = '/var/log/postgresql'
master_1  | ‣ Setting postgresql.conf parameter: log_filename = 'postgresql-9.4-main.log'
master_1  | ‣ Setting postgresql.conf parameter: ssl = 'off'
master_1  | Creating replication user: repl
master_1  | Starting PostgreSQL 9.4...
master_1  | LOG:  database system was shut down at 2016-08-08 06:17:27 UTC
master_1  | LOG:  MultiXact member wraparound protections are now enabled
master_1  | LOG:  autovacuum launcher started
master_1  | LOG:  database system is ready to accept connections
slave_1   | Waiting for master to accept connections (60s timeout)...
slave_1   | Replicating initial data from master...
master_1  | FATAL:  number of requested standby connections exceeds max_wal_senders (currently 0)
slave_1   | pg_basebackup: could not connect to server: FATAL:  number of requested standby connections exceeds max_wal_senders (currently 0)
slave_1   |
pgtest_slave_1 exited with code 1

Running the 9.4-22 compose file directly will make replication work out of the box, so this seems connected to the incremental upgrades from 9.4.

[pascalandy]

+1

I used

docker run --name postgresql-backup -it --rm \
  --link postgresql-master:master \
  --env 'REPLICATION_MODE=backup' --env 'REPLICATION_SSLMODE=prefer' \
  --env 'REPLICATION_HOST=master' --env 'REPLICATION_PORT=5432'  \
  --env 'REPLICATION_USER=repluser' --env 'REPLICATION_PASS=repluserpass' \
  --volume /srv/docker/backups/postgresql.$(date +%Y%m%d%H%M%S):/var/lib/postgresql \
  sameersbn/postgresql:9.5-1

I see a new folder getting created but it's empty.

---

Also about:

Restoring the backup involves starting a container with the data in /srv/docker/backups/postgresql.XXXXXXXXXXXX.

I'm confused here. If I relaunch the container it might take the a backup while I don't want to ?

[jokogr]

I had the same issue and solved it with the following:

docker exec -it postgresql bash
source ${PG_APP_HOME}/functions
configure_hot_standby

@pascalandy no, you have to change the --volume parameter to point to your backup directory