Latest release is not working

[Holt59]

Release 5.32.x is not working.

app fails to start because postgres 10+ is required but the Dockerfile for db specifies 9.4-alpine.

[adatamonk]

+1 seeing the same issue

{"level":"error","ts":1614369418.4518497,"caller":"commands/server.go:77","msg":"cannot create store: minimum required postgres version is 10.0; found 9.426"} Error: cannot create store: minimum required postgres version is 10.0; found 9.426

this is using for the DB : image: mattermost/mattermost-prod-db
and this for the app: image: mattermost/mattermost-team-edition

[ouafnico]

Hi
A lot of discussion about this postgresql version problem here: #489

[adatamonk]

Ya but at leased new setups should not be broken - currently a brand new setup is not working either

[coltoneshaw]

@adatamonk & @Holt59 - If you're using a brand new setup, you'll have to edit the file /db/Dockerfile. Below is what I have working on my test instance. Note the postgres version is now 13.2-alpine and the python-dev package is now python3-dev.

FROM postgres:13.2-alpine

ENV DEFAULT_TIMEZONE UTC

# Install some packages to use WAL
RUN echo "azure<5.0.0" > pip-constraints.txt
RUN apk add --no-cache \
      build-base \
      curl \
      libc6-compat \
      libffi-dev \
      linux-headers \
      python3-dev \
      py-pip \
      py-cryptography \
      pv \
      libressl-dev \
    && pip install --upgrade pip \
    && pip --no-cache-dir install -c pip-constraints.txt 'wal-e<1.0.0' envdir \
    && rm -rf /var/cache/apk/* /tmp/* /var/tmp/*

# Add wale script
COPY setup-wale.sh /docker-entrypoint-initdb.d/

#Healthcheck to make sure container is ready
HEALTHCHECK CMD pg_isready -U $POSTGRES_USER -d $POSTGRES_DB || exit 1

# Add and configure entrypoint and command
COPY entrypoint.sh /
ENTRYPOINT ["/entrypoint.sh"]
CMD ["postgres"]

VOLUME ["/var/run/postgresql", "/usr/share/postgresql/", "/var/lib/postgresql/data", "/tmp", "/etc/wal-e.d/env"]

This will not work directly for existing installs as you'll have to migrate your database to a new version

[adatamonk]

thanks @coltoneshaw - this is helpful. But do we know if the offical docker hub images are being fixed? it's easier to deploy those directly

[coltoneshaw]

@adatamonk - if you're using an existing install you can check out my comment here.

For a fresh install, I would think the version in the docker hub images will be updated soon but I'm not sure of an ETA for that. I think it's a bit more complex than just swapping the versions since it could break quite a few existing installs if not careful.

[shurrman]

gets more and more outdated

[tweibert]

Ran into the same problem today when upgrading from 5.31 to 5.33.

PostgreSQL Dockerfile was easily fixed to use PostgreSQL 13 - kudos to @coltoneshaw!

Here's how I did the cluster upgrade using the image from tianon/postgres-upgrade. This is all done inside the directory used for the /var/lib/postgresql volume.

First thing, prepare the directory layout expected by the upgrade scripts: move the existing data directory into a new 9.4 directory, and also create an empty directory 13 for the target:

$ mkdir 9.4
$ mv data 9.4
$ mkdir 13

Then run the upgrade script:

$ docker run --rm -e PGUSER=mmuser -e POSTGRES_INITDB_ARGS="-U mmuser" -v "$PWD":/var/lib/postgresql "tianon/postgres-upgrade:9.4-to-13"

Note the use of the PGUSER=mmuser and POSTGRES_INITDB_ARGS="-U mmuser" env variables here, as Mattermost is using mmuser as the superuser name, rather than the default postgres.

If all goes well, you should see:

Upgrade Complete
----------------
Optimizer statistics are not transferred by pg_upgrade so,
once you start the new server, consider running:
    ./analyze_new_cluster.sh

Running this script will delete the old cluster's data files:
    ./delete_old_cluster.sh

The pg_hba.conf file is not migrated automatically, so we copy it over manually:

$ cp 9.4/data/pg_hba.conf 13/data/pg_hba.conf

Then it's time to move the new data folder back to where it belongs:

$ mv 13/data .

Now start the Mattermost db and app containers. If everything is working, you can remove the 9.4 directory and the leftover log files and scripts from the upgrade-process.

This worked for me - use at your own risk, and remember to make backups!

[avoulk]

I simply used the script of @coltoneshaw and it worked like a charm!

[adatamonk]

i hope they can fix up the docker hub version though so we don't have to deploy this way....

[shurrman]

Do not try to upgrade minor 5.31 to 5.33, it will brake your MTM service