Upgrade to Pro
— share decks privately, control downloads, hide ads and more …
Speaker Deck
Features
Speaker Deck
PRO
Sign in
Sign up for free
Search
Search
Making GitLab Faster
Search
Yorick Peterse
June 04, 2016
Programming
2
450
Making GitLab Faster
Talk given at RubyC 2016
Yorick Peterse
June 04, 2016
Tweet
Share
More Decks by Yorick Peterse
See All by Yorick Peterse
Garbage Collection Crash Course
yorickpeterse
1
370
Rubinius & The Eternal Yak
yorickpeterse
1
250
Oga
yorickpeterse
3
190
Parsing for Humans
yorickpeterse
2
92
Other Decks in Programming
See All in Programming
メンテが命: PHPフレームワークのコンテナ化とアップグレード戦略
shunta27
0
310
責務と認知負荷を整える! 抽象レベルを意識した関心の分離
yahiru
8
1.4k
AWS Step Functions は CDK で書こう!
konokenj
4
480
Swift Testingのモチベを上げたい
stoticdev
2
130
技術を改善し続ける
gumioji
0
140
未経験でSRE、はじめました! 組織を支える役割と軌跡
curekoshimizu
1
170
Learning Kotlin with detekt
inouehi
1
150
「個人開発マネタイズ大全」が教えてくれたこと
bani24884
1
260
負債になりにくいCSSをデザイナとつくるには?
fsubal
10
2.6k
1年目の私に伝えたい!テストコードを怖がらなくなるためのヒント/Tips for not being afraid of test code
push_gawa
1
630
How mixi2 Uses TiDB for SNS Scalability and Performance
kanmo
41
16k
Honoをフロントエンドで使う 3つのやり方
yusukebe
7
3.6k
Featured
See All Featured
Practical Orchestrator
shlominoach
186
10k
The Cult of Friendly URLs
andyhume
78
6.2k
Evolution of real-time – Irina Nazarova, EuRuKo, 2024
irinanazarova
6
570
Designing Experiences People Love
moore
140
23k
[RailsConf 2023] Rails as a piece of cake
palkan
53
5.3k
Why You Should Never Use an ORM
jnunemaker
PRO
55
9.2k
The Pragmatic Product Professional
lauravandoore
32
6.4k
How to train your dragon (web standard)
notwaldorf
91
5.9k
GraphQLの誤解/rethinking-graphql
sonatard
68
10k
What’s in a name? Adding method to the madness
productmarketing
PRO
22
3.3k
Done Done
chrislema
182
16k
Adopting Sorbet at Scale
ufuk
74
9.2k
Transcript
Making GitLab Faster
@yorickpeterse
[email protected]
GitLab Code, test, and deploy together
GitLab.com Statistics Users 544 306 Projects 955 340 Requests per
minute ~60 000 Storage ~30 TB
Response Timings Mean 350 ms 95th Percentile 1.23 s 99th
Percentile 3.17 s (backend only)
GitLab is slow, GitLab.com even more so
None
GitLab Problems
Limited production monitoring New Relic only on 1 out of
40 servers
No good development tools Only a collection of random Gems
No performance guides No guides on good/bad patterns, snippets to
use, workflow, etc
Lack of specialists Nobody was specifically focusing on performance
Badly Performing Code N+1 queries, inefficient algorithms, loading lots of
data into memory, etc
Solutions According to HackerNews
“Gitlab is one project that absolutely needs to be rebuilt
using Java. Both from a performance and a deployment perspective” https://news.ycombinator.com/item?id=11049717
“gogs was designed from the start to be fast. Nothing
but a rewrite would make GitLab as fast as gogs for basic stuff on small self-hosted servers.” https://news.ycombinator.com/item?id=11431410
Problem Solved! Let’s Rewrite GitLab in Java/Go/Haskell
Production Monitoring
Production monitoring is the first step to solving performance problems
GitLab’s Requirements 1. FOSS 2. Easy to set up 3.
Easy to query, visualize and add metrics 4. Low overhead 5. Support for percentiles
N-th percentile: N% of values are up to a given
value
“The 90th percentile is 1.5 seconds” means 90% completes in
1.5 seconds
GitLab Performance Monitoring
InfluxDB Time series DB with SQL-like query syntax
Grafana Time series visualization
Custom Ruby Code Used for measuring data, built from scratch
Transaction Metrics Rails, Grape, and Sidekiq
Code Region Metrics For measuring custom regions of code
Background Sampler Object counts, memory usage, GC statistics
None
None
Development Tools
Sherlock Development-only performance analysis, re-using the GitLab UI
1. Request details (URL, timings, etc) 2. Query timings, backtraces,
and EXPLAIN ANALYZE output 3. Line profiling
None
None
None
Documentation
GitLab Performance Guide Documents patterns to use/avoid, workflow, etc
http://docs.gitlab.com/ce/development/performance.html or https://is.gd/gitlab_webscale
Application Code
The language is rarely the problem
The database is not a black box
Rampant abuse of “pluck”
users.where(id: projects.pluck(:user_id))
users.where(id: projects.select(:user_id))
Combining Query Results
user.projects + public_projects + internal_projects
user.projects UNION public_projects UNION internal_projects
UNION not properly supported in ActiveRecord/Arel :<
Lots of Git Operations
if repo.exists? ... end
if repo.master_branch? foo end
Git Tag Existence https://gitlab.com/gitlab-org/gitlab-ce/merge_requests/3983
if all_tags.include?(ref_name) => if tag_exists?(ref_name)
Markdown References https://gitlab.com/gitlab-org/gitlab-ce/merge_requests/3969
for text in texts for reference in text process(reference)
process(texts)
1. Grab all references in a single pass 2. Query
any rows using the references 3. Replace text using the queried data
Calling “url_helpers” https://gitlab.com/gitlab-org/gitlab-ce/merge_requests/3486
include app.routes.url_helpers
include Gitlab::Routing.url_helpers
def self.url_helpers @url_helpers ||= Gitlab::Application. routes.url_helpers end
Recap
Language/framework is usually only a problem when the application is
already fully optimized.
Take advantage of your database’s features
There is no silver bullet to fixing performance problems, it
simply takes time, effort, and patience.
Keep an eye on GitLab 8.9 for many performance improvements
Questions?