mahemoff

iTunes categories in most popular feeds at https://player.fm/featured/programming

Technology
Technology,Tech News,Technology,Software How-To
Technology
Technology,Software How-To
Technology,Tech News,Technology,Software How-To
Technology
Technology,Tech News,Education,Education Technology,Technology,Software How-To

mahemoff

# ruby 2.3.1p112 MacOS on MBP Retina late 2013
# Benchmarking some techniques at:
# https://stackoverflow.com/questions/12911869/is-there-a-faster-way-to-generate-random-string-in-ruby
# https://stackoverflow.com/questions/88311/how-to-generate-a-random-string-in-ruby

# Results show Object.hash is easily the fastest and shuffle is the slowest (as well as not being great as it doesn't support repeating characters)
# SecureRandom is a good compromise, especially if the string is needed for security purposes. It's 5 tims slower than Object.hash, but Object.hash
# is an int that would need converting to string to make an equivalent, smaller, representation.

[35] pry(main)> all=[]; Benchmark.measure { 10000.times { all << ('a'..'z').to_a.shuffle[0,8].join} }

mahemoff

Put database in read-only mode and capture master log position:

mysqL> FLUSH TABLES WITH READ LOCK; SET GLOBAL read_only = ON; SHOW MASTER STATUS;`

Copy entire mysql data folder:

bash> cp -r /var/lib/mysql /home/me/prod_varlibmysql20170715 ; date`

When copy is complete, immediately turn off read-only mode:

mahemoff

[<Linode api_id=499946, label='psaa'>]
[   {   u'AVAIL': {   u'10': 500,
                      u'11': 500,
                      u'2': 500,
                      u'3': 500,
                      u'4': 500,
                      u'6': 500,
                      u'7': 500,
                      u'8': 500,
                      u'9': 500},

mahemoff

The following MySQL command will show you long-running processes (exceeding 5 seconds in this case).

select * from information_schema.processlist where command <> 'sleep' and time > 5;

You can run it from console to view those happening now, or you could make a script to poll it periodically, with something like:

echo "select * from information_schema.processlist where command <> 'sleep' and time > 5;" | mysql > long-running.log

(See also https://serverfault.com/questions/416380/log-slow-queries-killed-before-completion/858093)

mahemoff

Class SidekiqUtils

  def self.queues
    ::Sidekiq::Stats.new.queues.keys.map { |name| ::Sidekiq::Queue.new(name) }
  end
  
   def self.find_queue(name) 
     self.queues.find { |q| q.name==name.to_s }
   end
  

mahemoff

- name: Backup original my.cnf
  copy: remote_src=true src=/etc/mysql/my.cnf dest=/tmp/my.cnf.recent

- name : Update my.cnf
  template: src=my.cnf.j2 dest=/etc/mysql/my.cnf owner=mysql mode=0644

# diff returns error code if different, so we ignore "errors"
- name: Check if my.cnf changed
  command: diff /etc/mysql/my.cnf /tmp/my.cnf.recent
  ignore_errors: true

mahemoff

RATIONALE

It's important to have useful test data during development and debugging, resembling real-world data. It can be cumbersome to setup and maintain though. This is a mini-framework to help with seeding in Rails using a Domain Specific Language approach.

DESIGN APPROACH

• I got tired of maintaining seeds.rb idempotently. Too dangerous to run this directly into production and overall, cumbersome to make it idempotent (ie work whether or not the objects already exist). So I decided to just make something that's just used for development, starting from an empty DB each time. I'm now maintaining a separate folder of individual data migrations to be performed in production. There's also separate fixtures for testing (separate data than this for performance reasons, and because tests need different kinds of data).
• It's modular enough for parts to be invoked during testing.
• The modularity is also part of a domain-specific language approach to seeding, as can be seen in the examples.

mahemoff

Sometimes it gets into inconsistent state. For testing purposes, we don't want to debug it, we just want to start again.

1. kill elasticsearch if it's running
2. brew info elasticsearch
3. Look for a line like `Data: /usr/local/var/elasticsearch/elasticsearch_foo/
4. rm -fr /usr/local/var/elasticsearch/elasticsearch_foo
5. restart elasticsearch and enjoy your tabula rasa

mahemoff

Something I noticed after settiing use_transactional_tests = false is that fixture files need to be present even if they don't have any data. An empty yml file means the model will be reset after every test run, i.e. you won't have side effects from creating new instances in a test.

It feels off to see an empty file (or just commented) and keep it in the project, but you need to do it.