Discourse Bad Gateway after reboot

My server runs on in a Virtual Machine hosted by one of the major cloud providers.
I successfully installed discourse on it and it’s been running fine for the last month.
Today, I decided to change the specifications of my VM back to its original configuration(*) and rebooted. Upon startup, whilst everything else on my server is running fine, I’m getting a 502 Bad Gateway error when I attempt to access the discourse forum. Thinking the docker instance had not automatically started up, I SSHed into my server and ran ./launcher start app but got a message saying I had insufficient space remaining (5GB available), so I ran df -h which tells me I actually have 14GB available. So I ran ./launcher start app again, but this time I got a warning that docker was going to download stuff and to be patient. After some processing, I got the message Nothing to do, your container has already started!. However, my attempts to access the forum still returned 502 Bad Gateway.
After consulting this forum here, I decided to run ./launcher rebuild app and got the following errors, something to do with PostgreSQL:

    user@host:[16:48]:/var/discourse# ./launcher rebuild app
    Ensuring launcher is up to date
    Fetching origin
    Launcher is up-to-date
    Stopping old container
    + /usr/bin/docker stop -t 60 app
    app
    cd /pups && git pull && /pups/bin/pups --stdin
    Already up to date.
    I, [2020-07-01T07:19:42.821347 #1]  INFO -- : Loading --stdin
    I, [2020-07-01T07:19:42.831806 #1]  INFO -- : > locale-gen $LANG && update-locale
    I, [2020-07-01T07:19:42.879007 #1]  INFO -- : Generating locales (this might take a while)...
    Generation complete.
    
    I, [2020-07-01T07:19:42.879431 #1]  INFO -- : > mkdir -p /shared/postgres_run
    I, [2020-07-01T07:19:42.885054 #1]  INFO -- :
    I, [2020-07-01T07:19:42.885734 #1]  INFO -- : > chown postgres:postgres /shared/postgres_run
    I, [2020-07-01T07:19:42.891655 #1]  INFO -- :
    I, [2020-07-01T07:19:42.892269 #1]  INFO -- : > chmod 775 /shared/postgres_run
    I, [2020-07-01T07:19:42.898103 #1]  INFO -- :
    I, [2020-07-01T07:19:42.898942 #1]  INFO -- : > rm -fr /var/run/postgresql
    I, [2020-07-01T07:19:42.905607 #1]  INFO -- :
    I, [2020-07-01T07:19:42.906463 #1]  INFO -- : > ln -s /shared/postgres_run /var/run/postgresql
    I, [2020-07-01T07:19:42.912617 #1]  INFO -- :
    I, [2020-07-01T07:19:42.913233 #1]  INFO -- : > socat /dev/null UNIX-CONNECT:/shared/postgres_run/.s.PGSQL.5432 || exit 0 && echo postgres already running stop container ; exit 1
    2020/07/01 07:19:42 socat[26] E connect(6, AF=1 "/shared/postgres_run/.s.PGSQL.5432", 36): No such file or directory
    I, [2020-07-01T07:19:42.925688 #1]  INFO -- :
    I, [2020-07-01T07:19:42.926081 #1]  INFO -- : > rm -fr /shared/postgres_run/.s*
    I, [2020-07-01T07:19:42.931174 #1]  INFO -- :
    I, [2020-07-01T07:19:42.931649 #1]  INFO -- : > rm -fr /shared/postgres_run/*.pid
    I, [2020-07-01T07:19:42.938152 #1]  INFO -- :
    I, [2020-07-01T07:19:42.938850 #1]  INFO -- : > mkdir -p /shared/postgres_run/12-main.pg_stat_tmp
    I, [2020-07-01T07:19:42.943575 #1]  INFO -- :
    I, [2020-07-01T07:19:42.944331 #1]  INFO -- : > chown postgres:postgres /shared/postgres_run/12-main.pg_stat_tmp
    I, [2020-07-01T07:19:42.949159 #1]  INFO -- :
    I, [2020-07-01T07:19:42.961190 #1]  INFO -- : File > /etc/service/postgres/run  chmod: +x  chown:
    I, [2020-07-01T07:19:42.973345 #1]  INFO -- : File > /etc/service/postgres/log/run  chmod: +x  chown:
    I, [2020-07-01T07:19:42.983929 #1]  INFO -- : File > /etc/runit/3.d/99-postgres  chmod: +x  chown:
    I, [2020-07-01T07:19:42.994843 #1]  INFO -- : File > /root/upgrade_postgres  chmod: +x  chown:
    I, [2020-07-01T07:19:42.995487 #1]  INFO -- : > chown -R root /var/lib/postgresql/12/main
    I, [2020-07-01T07:19:44.012812 #1]  INFO -- :
    I, [2020-07-01T07:19:44.013656 #1]  INFO -- : > [ ! -e /shared/postgres_data ] && install -d -m 0755 -o postgres -g postgres /shared/postgres_data && sudo -E -u postgres /usr/lib/postgresql/12/bin/initdb -D /shared/postgres_data || exit 0
    I, [2020-07-01T07:19:44.019545 #1]  INFO -- :
    I, [2020-07-01T07:19:44.019872 #1]  INFO -- : > chown -R postgres:postgres /shared/postgres_data
    I, [2020-07-01T07:19:44.064432 #1]  INFO -- :
    I, [2020-07-01T07:19:44.065186 #1]  INFO -- : > chown -R postgres:postgres /var/run/postgresql
    I, [2020-07-01T07:19:44.071385 #1]  INFO -- :
    I, [2020-07-01T07:19:44.072196 #1]  INFO -- : > /root/upgrade_postgres
    I, [2020-07-01T07:19:44.084004 #1]  INFO -- :
    I, [2020-07-01T07:19:44.084662 #1]  INFO -- : > rm /root/upgrade_postgres
    I, [2020-07-01T07:19:44.090399 #1]  INFO -- :
    I, [2020-07-01T07:19:44.092280 #1]  INFO -- : Replacing data_directory = '/var/lib/postgresql/12/main' with data_directory = '/shared/postgres_data' in /etc/postgresql/12/main/postgresql.conf
    I, [2020-07-01T07:19:44.093969 #1]  INFO -- : Replacing (?-mix:#?listen_addresses *=.*) with listen_addresses = '*' in /etc/postgresql/12/main/postgresql.conf
    I, [2020-07-01T07:19:44.095204 #1]  INFO -- : Replacing (?-mix:#?synchronous_commit *=.*) with synchronous_commit = $db_synchronous_commit in /etc/postgresql/12/main/postgresql.conf
    I, [2020-07-01T07:19:44.095937 #1]  INFO -- : Replacing (?-mix:#?shared_buffers *=.*) with shared_buffers = $db_shared_buffers in /etc/postgresql/12/main/postgresql.conf
    I, [2020-07-01T07:19:44.096695 #1]  INFO -- : Replacing (?-mix:#?work_mem *=.*) with work_mem = $db_work_mem in /etc/postgresql/12/main/postgresql.conf
    I, [2020-07-01T07:19:44.097554 #1]  INFO -- : Replacing (?-mix:#?default_text_search_config *=.*) with default_text_search_config = '$db_default_text_search_config' in /etc/postgresql/12/main/postgresql.conf
    I, [2020-07-01T07:19:44.101971 #1]  INFO -- : > install -d -m 0755 -o postgres -g postgres /shared/postgres_backup
    I, [2020-07-01T07:19:44.112672 #1]  INFO -- :
    I, [2020-07-01T07:19:44.113831 #1]  INFO -- : Replacing (?-mix:#?max_wal_senders *=.*) with max_wal_senders = $db_max_wal_senders in /etc/postgresql/12/main/postgresql.conf
    I, [2020-07-01T07:19:44.114973 #1]  INFO -- : Replacing (?-mix:#?wal_level *=.*) with wal_level = $db_wal_level in /etc/postgresql/12/main/postgresql.conf
    I, [2020-07-01T07:19:44.116047 #1]  INFO -- : Replacing (?-mix:#?checkpoint_segments *=.*) with checkpoint_segments = $db_checkpoint_segments in /etc/postgresql/12/main/postgresql.conf
    I, [2020-07-01T07:19:44.117033 #1]  INFO -- : Replacing (?-mix:#?logging_collector *=.*) with logging_collector = $db_logging_collector in /etc/postgresql/12/main/postgresql.conf
    I, [2020-07-01T07:19:44.118051 #1]  INFO -- : Replacing (?-mix:#?log_min_duration_statement *=.*) with log_min_duration_statement = $db_log_min_duration_statement in /etc/postgresql/12/main/postgresql.conf
    I, [2020-07-01T07:19:44.119352 #1]  INFO -- : Replacing (?-mix:^#local +replication +postgres +peer$) with local replication postgres  peer in /etc/postgresql/12/main/pg_hba.conf
    I, [2020-07-01T07:19:44.120299 #1]  INFO -- : Replacing (?-mix:^host.*all.*all.*127.*$) with host all all 0.0.0.0/0 md5 in /etc/postgresql/12/main/pg_hba.conf
    I, [2020-07-01T07:19:44.121038 #1]  INFO -- : > HOME=/var/lib/postgresql USER=postgres exec chpst -u postgres:postgres:ssl-cert -U postgres:postgres:ssl-cert /usr/lib/postgresql/12/bin/postmaster -D /etc/postgresql/12/main
    I, [2020-07-01T07:19:44.126334 #1]  INFO -- : > sleep 5
    2020-07-01 07:19:44.157 UTC [49] LOG:  starting PostgreSQL 12.2 (Debian 12.2-2.pgdg100+1) on x86_64-pc-linux-gnu, compiled by gcc (Debian 8.3.0-6) 8.3.0, 64-bit
    2020-07-01 07:19:44.158 UTC [49] LOG:  listening on IPv4 address "0.0.0.0", port 5432
    2020-07-01 07:19:44.158 UTC [49] LOG:  listening on IPv6 address "::", port 5432
    2020-07-01 07:19:44.161 UTC [49] LOG:  listening on Unix socket "/var/run/postgresql/.s.PGSQL.5432"
    2020-07-01 07:19:44.162 UTC [49] FATAL:  could not map anonymous shared memory: Cannot allocate memory
    2020-07-01 07:19:44.162 UTC [49] HINT:  This error usually means that PostgreSQL's request for a shared memory segment exceeded available memory, swap space, or huge pages. To reduce the request size (currently 4423172096 bytes), reduce PostgreSQL's shared memory usage, perhaps by reducing shared_buffers or max_connections.
    2020-07-01 07:19:44.162 UTC [49] LOG:  database system is shut down
    I, [2020-07-01T07:19:49.141762 #1]  INFO -- :
    I, [2020-07-01T07:19:49.142221 #1]  INFO -- : > su postgres -c 'createdb discourse' || true
    createdb: error: could not connect to database template1: could not connect to server: No such file or directory
        Is the server running locally and accepting
        connections on Unix domain socket "/var/run/postgresql/.s.PGSQL.5432"?
    I, [2020-07-01T07:19:49.227852 #1]  INFO -- :
    I, [2020-07-01T07:19:49.228226 #1]  INFO -- : > su postgres -c 'psql discourse -c "create user discourse;"' || true
    psql: error: could not connect to server: could not connect to server: No such file or directory
        Is the server running locally and accepting
        connections on Unix domain socket "/var/run/postgresql/.s.PGSQL.5432"?
    I, [2020-07-01T07:19:49.330486 #1]  INFO -- :
    I, [2020-07-01T07:19:49.330822 #1]  INFO -- : > su postgres -c 'psql discourse -c "grant all privileges on database discourse to discourse;"' || true
    psql: error: could not connect to server: could not connect to server: No such file or directory
        Is the server running locally and accepting
        connections on Unix domain socket "/var/run/postgresql/.s.PGSQL.5432"?
    I, [2020-07-01T07:19:49.425970 #1]  INFO -- :
    I, [2020-07-01T07:19:49.426356 #1]  INFO -- : > su postgres -c 'psql discourse -c "alter schema public owner to discourse;"'
    psql: error: could not connect to server: could not connect to server: No such file or directory
        Is the server running locally and accepting
        connections on Unix domain socket "/var/run/postgresql/.s.PGSQL.5432"?
    I, [2020-07-01T07:19:49.506638 #1]  INFO -- :
    I, [2020-07-01T07:19:49.507202 #1]  INFO -- : Terminating async processes
    
    
    FAILED
    --------------------
    Pups::ExecError: su postgres -c 'psql discourse -c "alter schema public owner to discourse;"' failed with return #<Process::Status: pid 75 exit 2>
    Location of failure: /pups/lib/pups/exec_command.rb:112:in `spawn'
    exec failed with the params "su postgres -c 'psql $db_name -c \"alter schema public owner to $db_user;\"'"
    eb41679f76cd749ccd8c84a7543365d093619b80df6fc6750b9349fb63565fa1
    ** FAILED TO BOOTSTRAP ** please scroll up and look for earlier error messages, there may be more than one.
    ./discourse-doctor may help diagnose the problem.
    user@host:[17:19]:/var/discourse#

Strangely, in spite of the errors above, running ./launcher start app produces no errors:

starting up existing container
+ /usr/bin/docker start app
app

With the instance running, I tried using ./launcher enter app to get into the container. (In my humble opinion the available tools in the container are very poor (yes I’m a nano user and like to have various aliases mapped; e.g. ll). I’m not able to find the physical path to the folders within the docker instance (as I would like to download them using a FTP client).

In /var/log/nginx/error.log I have the following error entry for each time I refresh my browser:

2020/07/01 07:44:16 [error] 646#646: *3 connect() failed (111: Connection refused) while connecting to upstream, client: xxx.xx.0.1, server: _, request: "GET / HTTP/1.1", upstream: "http://127.0.0.1:3000/", host: "discourse.myDomain.com"

What could be the cause of my problem? Why is Postgre suddenly not functioning?

(*) A week after installing Discourse, I upgraded my server with more CPUs and memory. I needed to do this to run a video conference I hosted. With the conference done, I’ve gone back to my normal configuration. Note that I did not change the disk sizes at any point during the specification changes.

1 Like

This is because your current rebuild of the container failed; and you are starting a prior version of your app. This is the normal behavior. When a rebuild is not successful, the original container is not deleted (generally speaking) and the original image also remains available.

Regardingyour PG problem, you will need to provide th team more details of your app and container configuration, to get the best support.

2 Likes

@neounix : Thanks.

I’m new to hosting a discourse forum so I’m not sure where to look and what to lookout for. I have an essentially vanilla install with no plugins or other modifications. I have some variables defined in app.yml and I’m using my existing Apache2 daemon as a reverse proxy to forward Discourse traffic, via a separate virtualhost, to the localhost port I’ve configured Discourse to listen to.
Could you elaborate on what information would be useful? Is there a resource I could read that would help me debug my situation?

The core error is in the log file run above.

2020-07-01 07:19:44.162 UTC [49] FATAL:  could not map anonymous shared memory: Cannot allocate memory

2020-07-01 07:19:44.162 UTC [49] HINT:  This error usually means that PostgreSQL's request for a shared memory segment exceeded available memory, swap space, or huge pages. To reduce the request size (currently 4423172096 bytes), reduce PostgreSQL's shared memory usage, perhaps by reducing shared_buffers or max_connections.
2 Likes

I saw that error but I haven’t made any changes in the app.yml.
Where can I reduce the shared_buffers or max_connections, they are not in app.yml? app.yml only has a parameter db_shared_buffers but that is set to the default value “4096MB” as it has been all along (before and after I increase the server’s memory).

You might consider posting your stats related to memory.

For example, on Linux:

$ free -m
              total        used        free      shared  buff/cache   available
Mem:          64299       12955        9678         361       41664       50265
Swap:          7807          69        7738

and for docker stats, post the output of

docker stats

etc.

The error is related to a lack of memory.

2 Likes

The server mem stats are:

              total        used        free      shared  buff/cache   available
Mem:           3951        2236         414          86        1299        1308
Swap:           511         415          96

mem stats after enter app are:

              total        used        free      shared  buff/cache   available
Mem:           3951        2363         321          86        1266        1215
Swap:           511         415          96

Running docker stats > output.txt produced:

        CONTAINER ID        NAME                CPU %               MEM USAGE / LIMIT    MEM %               NET I/O             BLOCK I/O           PIDS
       ca4c5f37894c        app                 15.86%              6.48MiB / 3.859GiB   0.16%               20.3kB / 12.6kB     0B / 0B             25
        CONTAINER ID        NAME                CPU %               MEM USAGE / LIMIT    MEM %               NET I/O             BLOCK I/O           PIDS
       ca4c5f37894c        app                 15.86%              6.48MiB / 3.859GiB   0.16%               20.3kB / 12.6kB     0B / 0B             25
        CONTAINER ID        NAME                CPU %               MEM USAGE / LIMIT     MEM %               NET I/O             BLOCK I/O           PIDS
       ca4c5f37894c        app                 2.83%               6.539MiB / 3.859GiB   0.17%               20.3kB / 12.6kB     0B / 0B             25
        CONTAINER ID        NAME                CPU %               MEM USAGE / LIMIT     MEM %               NET I/O             BLOCK I/O           PIDS
       ca4c5f37894c        app                 2.83%               6.539MiB / 3.859GiB   0.17%               20.3kB / 12.6kB     0B / 0B             25
        CONTAINER ID        NAME                CPU %               MEM USAGE / LIMIT     MEM %               NET I/O             BLOCK I/O           PIDS
       ca4c5f37894c        app                 3.30%               6.477MiB / 3.859GiB   0.16%               20.3kB / 12.6kB     0B / 0B             25
        CONTAINER ID        NAME                CPU %               MEM USAGE / LIMIT     MEM %               NET I/O             BLOCK I/O           PIDS
       ca4c5f37894c        app                 3.30%               6.477MiB / 3.859GiB   0.16%               20.3kB / 12.6kB     0B / 0B             25
        CONTAINER ID        NAME                CPU %               MEM USAGE / LIMIT     MEM %               NET I/O             BLOCK I/O           PIDS
       ca4c5f37894c        app                 2.45%               6.535MiB / 3.859GiB   0.17%               20.3kB / 12.6kB     0B / 0B             25
        CONTAINER ID        NAME                CPU %               MEM USAGE / LIMIT     MEM %               NET I/O             BLOCK I/O           PIDS
       ca4c5f37894c        app                 2.45%               6.535MiB / 3.859GiB   0.17%               20.3kB / 12.6kB     0B / 0B             25
2 Likes

Hi @nap

You can reclaim a lot of memory by stopping and then removing all those old app containers.

For example:

docker stop <container_id>
docker rm <container_id>

Assuming they are not in use?

If they are all in use, then you should consider increasing the memory for this server above 4GB; maybe go for 8GB :slight_smile:

3 Likes

I stopped the app ./launcher stop app and then re-ran docker stats. There were no containers listed.
Unfortunately increasing the memory means paying more. The frustrating thing at the moment is that it was working last month with 4GB.

1 Like

And I’m not even able to rebuild at the moment, which can’t consume that much memory.

Without the container running, mem stats are:

              total        used        free      shared  buff/cache   available
Mem:           3951        2207         169          91        1574        1332
Swap:           511         446          65

I have a few interesting directories in ./var/lib/docker/overlay2/

e3e6cdfcc62c2e0b68ec91efxxxxx6c69212c95b5070f7b6b84e97edcb473ea2
64a04d1b97a18f51a5fdc536xxxxxf9473de0c2ccd1a2cc0d62e830164b5f2d8
355303c6af7bebff1163195c5xxxxx8fd1de6333e39adbcb573c7365673b6c85

Can I delete these?

Right.

I see. I was busy working on another task and did not notice that your output showed the stats for the same container and not multiple ones.

What does free -m tell you now that your container is not running?

I think the 4GB RAM is OK for the one container, for sure.

No.

Don’t delete those docker files.

The issue, from the error message, is related to your Discourse PG 12 configuration. I’m not sure how to address that, because tweaking the PG 12 conf file for Discourse is not supported, don’t think.

The meta top guns will have better suggestions than me, especially the professional hosting teams.

2 Likes

What you’re saying is that this is internal to the files within the docker configuration? And modifying it manually will cause problems once the container is started or updated?

@nap

If you do a google search for the error message above (in quotes), you will find a number of directly relevant discussions on this exact PostgreSQL error message.

Hope this helps.

3 Likes

After you did that did you re-run ./discourse-setup or modify the memory settings in the app.yml by hand? What are db_shared_buffers, unicorn_workers and db_work_mem?

Except you’re running behind a reverse proxy, which makes things more complicated. It’s not clear that the reverse proxy is at fault here, but it does complicate things.

Do you have multiple partitions? Could it be that the partition where docker creates images is full?

4 Likes

@pfaffman : Thanks for having a look.

No, all I did was add a series of variable definitions that pertain to the site name and the use of tags

db_shared_buffers is “4096MB”
unicorn_workers is 8
db_work_mem is commented out

I have one main 40G partition (14GB free), 512MB swap, and an 8GB partition for backups (not mounted).

It appears I have overcome the problem. Initially I tried reducing buffers to 2GB and workers to 4 but got the same error. I then reduced the buffers down to 1GB whereupon rebuild succeeded and the forum is now back up.

Thank you all!!

3 Likes