From Bitpost wiki
Jump to navigation Jump to search

Install system packages

  • Use this to install the appropriate system postgres, and libpqxx from source:

On Ubuntu, mh-install-postgres will currently install postgres 12 on Ubuntu 18.04, and configure it for you. It will give you notes at the end to help you finish installation.

On gentoo, we just need to unmask the postgresql-12.1 package. No need to set up a data folder though.

🌵 m@bitpost  [~] emacs /etc/portage/package.keywords
  # MDM postgres 12 is currently masked cmon
🌵 m@bitpost  [~] sudo emerge -DavuN =postgresql-12.1
    • The config files are here (but you won't need them):

TODO fold all this ^ into mh-install-postgres

Disabling multiple installs

There may be one or more system postgres'es already installed. That's fine, just make sure they don't start.

sudo emacs -nw /etc/postgresql/(tab-tab)
sudo emacs -nw /etc/postgresql/10/main/start.conf # change auto > manual
sudo emacs -nw /etc/postgresql/9.6/main/start.conf # change auto > manual
sudo service postgres(tab-tab to determine if there are more than one) stop

On Ubuntu, watch out for multiple installs causing the postgres port to change from the default. If you have two previous postgres versions, version 12 will use port 5432+1+1 = 5434. This is very confusing. You can see what postgres versions are running on what port in Ubuntu with this:


Start server

Use mh-start-postgres to start postgres (if the server is not already running):


Add a user + db

First, change the postgres user password:

sudo su postgres
\password postgres
(Enter new password...)

Use mh-add-postgres-db to add a user (aka role) and matching database. The user only has access to that db (best practice). This will prompt for pw.

mh-add-postgres-db [name]
# Add name+pw to private


Readable UNIX epoch

select to_timestamp((json->>'timestamp')::int) at time zone 'utc', * from stockquotes where (json->>'quote')::float < 0.0;

Yes it's that ridiculous. Yes it's really even worse than that.

Table sizes

More ridiculous postgres overengineered obfuscated NONSENSE:

SELECT nspname || '.' || relname AS "relation",
   pg_size_pretty(pg_total_relation_size(C.oid)) AS "total_size"
 FROM pg_class C
 LEFT JOIN pg_namespace N ON (N.oid = C.relnamespace)
 WHERE nspname NOT IN ('pg_catalog', 'information_schema')
   AND C.relkind <> 'i'
   AND nspname !~ '^pg_toast'
 ORDER BY pg_total_relation_size(C.oid) DESC;

Optionally add LIMIT 5 or whatever.


Read this carefully!

  • Add columns of type 'jsonb' (JSON binary).
  • Inject cells of any valid JSON format (big data!).
  • Then using special syntax, you can crack open the json inside a cell to query, create, update, delete:
   CREATE TABLE stocks (
     id integer NOT NULL,
     data jsonb
   INSERT INTO stocks VALUES (1, '{"symbol": "GOOG", "tags": ["hot", "trending"], "owned": true}, "quantity": 12');
   # returns count of data that HAS the field ingredients
   SELECT count(*) FROM stocks WHERE data ? 'tags';
   # ->> will return the property as integer or text (the parsed form of the attribute).
   SELECT * FROM stocks WHERE data->>'owned' = 'true';         # if data.owned == true
   # you may need to cooerce the type, sigh...
   SELECT * FROM stocks WHERE (data->>'quantity')::int > 5;
   # -> will return the attribute as a JSON object
   SELECT jsonb_array_elements_text(data->'tags') as tag FROM stocks WHERE id = 1;
   # returns the actual tag values, eg:
       (2 rows)

Field(s) update

So easy with v12 syntax:

UPDATE mytable SET json = json || '{"order_id": ""}';
update prefint set json = json || '{"value":1}' where json->>'name' = 'LoggingLevel';

Field(s) delete

UPDATE mytable SET json = json - 'custom_order_id';


Upsert example, from SO:

INSERT INTO accounts(id, json) VALUES ('1', '{##somejson##}'); 

INSERT INTO accounts(id, json) 
VALUES ('1', '{##somejson##}')
UPDATE SET json = '{##somejson##}'




  • I had bad data (huge string written to a json value) and found it by looking at the json length not the actual json (which was huge).
select id, length(json::text) from cycles;

Then I could clean up the rogue field with:

update cycles set json = json || '{"order_id": ""}';


Look at [at import] and [at export], they do the work.


  • use mh to install postgres and libpqxx to /usr/local
  • set up a linux user named "postgres"
# On gentoo:
sudo useradd postgres
# On ubuntu:
sudo adduser postgres
# Put the password in private
  • set up a data folder where postgres will put all its data and its log
# Standardize on [/home/m/development/postgres], but always consider symlinking to a large/redundant location.
# Here, we are choosing to use a larger RAID drive.
# WARNING NTFS will not let you chown a folder (you'd have to mount the whole thing as the target owner).  
# So on viper, I had to use the tiny primary drive directly, no symlink to ~m/bigdata.  Fuck off.
mkdir /spiceflow/softraid/development/postgres
cd ~/development && ln -s /spiceflow/softraid/development/postgres postgres
sudo chown -R postgres:postgres postgres /spiceflow/softraid/development/postgres
sudo chmod g+x /spiceflow/softraid/development # and any other parent that is missing this bit, so postgres user can cd to its folder

# We must init the data folder as a postgres "database cluster"
su postgres -c '/usr/local/bin/initdb -D /home/m/development/postgres/data'


To dump a database called mydb into a SQL-script file:

$ pg_dump mydb > db.sql

To reload such a script into a (freshly created) database named newdb:

$ psql -d newdb -f db.sql

See here for more.


You need to create two config files, postgresql.conf and pg_hba.conf.

There is a great configuration tool here to generate postgresql.conf. Example result:

# DB Version: 10
# OS Type: linux
# DB Type: web
# Total Memory (RAM): 16 GB
# CPUs num: 4
# Connections num: 1000
# Data Storage: hdd

max_connections = 1000
shared_buffers = 4GB
effective_cache_size = 12GB
maintenance_work_mem = 1GB
checkpoint_completion_target = 0.7
wal_buffers = 16MB
default_statistics_target = 100
random_page_cost = 4
effective_io_concurrency = 2
work_mem = 2097kB
min_wal_size = 1GB
max_wal_size = 2GB
max_worker_processes = 4
max_parallel_workers_per_gather = 2
max_parallel_workers = 4

You must provide the pg_hba.conf file to specify access control. Copy it from other installs. Make sure to add any host-to-host access you need.

Here is the location to put the config files when installed from source with [mh-install-postgres]: