Failure, exiting: The source cluster was not shut down cleanly

[stephane-klein]

First I stop my postgres 9.4 service instance, next:

$ tree -L 2
.
├── 12
│   └── data
├── 9.4
│   └── data
├── data
├── pg_upgrade_internal.log
├── pg_upgrade_server.log
└── pg_upgrade_utility.log
$ docker run --rm -v `pwd`/:/var/lib/postgresql tianon/postgres-upgrade:9.4-to-12
Performing Consistency Checks
-----------------------------
Checking cluster versions                                   ok

The source cluster was not shut down cleanly.
Failure, exiting

Can you help me? What is my mistake?

Best regards,
Stéphane

[stephane-klein]

This is how I fixed it:

docker run -it --rm -v `pwd`/:/var/lib/postgresql/ tianon/postgres-upgrade:9.4-to-12 bash
su postgres
$PGBINOLD/pg_ctl start -w -D $PGDATAOLD
$PGBINOLD/pg_ctl stop -w -D $PGDATAOLD
exit

Next I can start this with success:

docker run --rm -v `pwd`/:/var/lib/postgresql tianon/postgres-upgrade:9.4-to-12

[stephane-klein]

@tianon what to you think to add pg_ctl start + pg_ctl stop step before execute pg_upgrade to avoid The source cluster was not shut down cleanly error?

[harobed]

Same commands in one line:

docker run -it --rm -v `pwd`/:/var/lib/postgresql/ tianon/postgres-upgrade:9.4-to-12 su postgres -c "\$PGBINOLD/pg_ctl start -w -D \$PGDATAOLD ; \$PGBINOLD/pg_ctl stop -w -D \$PGDATAOLD; /usr/local/bin/; /usr/local/bin/docker-upgrade"

[stephane-klein]

Before:

docker run -it --rm -v `pwd`/:/var/lib/postgresql/ tianon/postgres-upgrade:9.4-to-12 su postgres -c "\$PGBINOLD/pg_ctl start -w -D \$PGDATAOLD ; \$PGBINOLD/pg_ctl stop -w -D \$PGDATAOLD; /usr/local/bin/; /usr/local/bin/docker-upgrade"

Don't forget to execute:

# chown 999 ./ -R

[snspinn]

I found that running docker exec --user postgres POSTGRES bash -c "pg_ctl stop" worked --> where POSTGRES is the name of the original (and running) postgres container. Just make sure that docker (or docker-compose) is not setup to restart your container on failure.

I wonder if the container/docker runs pg_ctl stop with the 'immediate' directive, i.e. pg_ctl stop -m i?