Kubernetes Cloud Native Indonesia Meetup - June 2024

Kubernetes Cloud Native Indonesia Meetup - June 2024

1. Kubernetes Cloud Native
Meetup June 2024

2. Agenda
●Kubernetes Community Update
●Talk #1 by Samir H., Training Manager @ Upcloud
●Talk #2 by Nicolas Julian, Performance engineer @
Jubelio
●Talk #3 by Irwan Shofwan, Cloud native enthusiast
●Q&A
●Photo Session

3. Update from
Kubernetes & Cloud Native Indonesia

4. Kubernetes Community Days
Indonesia 2024

5. CFP, April 2024 ~ 2025

https://events.linuxfoundation.org/about/calendar/

6. Speakers

7. Talk #1

8. UpCloud Kubernetes Service (UKS)
Samir Haliru
Training Manager

9. About me: Samir Haliru
●Joined UpCloud Support team 5 years ago.
●Technical training for new and existing
members of the Support team
●Used to work and live in Jakarta

10. UpCloud: Who We
Are
UpCloud is a leading European cloud service provider
with a global cloud platform spanning four continents.
We offer a reliable, GDPR-compliant cloud-native
product portfolio available from 13 different data
centres.
●10 000+ customers
●13 data centres globally
●24/7 customer service
upcloud.com

11. Today’s agenda
➔ A quick Introduction to Kubernetes
➔ UpCloud Kubernetes Service (UKS)
➔ Short demo

12. What is
Kubernetes?

13. What is Kubernetes?
●Open-source container orchestration platform.
●Developed by Google and now maintained by the Cloud
Native Computing Foundation (CNCF).
●Its primary purpose is to automate the deployment, scaling,
and management of containerised applications.

14. Benefits of Kubernetes?
●Automation
●Scalability
●Efficiency

15. UpCloud
Kubernetes Service
(UKS).

16. UpCloud Kubernetes
Service (UKS)
UKS is a fully serviced container orchestration system provided by
UpCloud. It easily automates the deployment, scaling and
management of container workloads.
UpCloud’s Managed Kubernetes is a part of CNCF’s Certified
Kubernetes Conformance Program.

17. UKS architecture

18. Some features of UKS

●Simplified cluster creation and management
●Intuitive dashboard and management tools
●Built-in support for high availability
●Compatibility with various tools

19. UKS demo.

20. Documentation available at
https://upcloud.com/docs/

Special promo code for event attendees
CNCFJKT0624
Tutorials and
documentation
Free trial and starting
credits
How to get started

21. Thank you.

22. Talk #2

23. Observe
Performance in
Kubernetes

24. Nicolas Julian
Sekarang bekerja sebagai “Kuli Server” di
Jubelio.
Latar belakang, Seorang warga sipil bisa yang
suka ngopi sambil sebats,…
Who ?
24

25. ◉You know your k8s cluster doing better.
◉Gain performance win.
◉More hate work of ministry of telecommunication.
What will you know by this
talk
25

26. 26
What does observe on mean? a. : to watch
carefully especially with attention to details or
behavior for the purpose of arriving at a judgment.

27. How to Observe (in k8s)
27
Kube components
○Kiali
○kubernetes/kube-state-metrics
○Kubernetes Dashboard, etc…
Apps inside k8s
○Datadog
○NewRelic
○APM by Elastic
○Otel, etc…

28. Common Observe Performance Problem
28

●Manual instrumentation not sometimes not covering blindspot
●High CPUs
●Network Overhead sending data to centralize Performance Monitoring

29. Blind Spot
●Kubernetes service
communication is based
on alot of network
instrumentation, that
create new blind spot for
performance. With e-bpf
agent siting in the kernel
watch every event, all
events is traceable.
Resources
●In term of resources,
eBPF is very lightweight
compare to others.
Why e-BPF Agent?
29

30. What ebpf agent suite our need (Jubelio)
30
●OpenSources.
●75% our traffic is egress.
●Need to trace the TLS Encrypted outgoing
http packet.
●Low footprint cpu/memory.
●Low overhead network traffic.
●Auto-Instrumentation.

31. As Far I Know
31
●https://github.com/inspektor-gadget/ (Debuging)

●https://github.com/pixie-io/pixie (GUI)

32. Inspektor-gadget
32

33. Pixie
33

34. How we utilise the tools so far – Debug not efficient resolving DNS
mechanism

34
Resolve with
small changes in
/etc/resolve.conf

Update Value
ndots:1

Speed up 10x
better.

35. How we utilise the tools so far – Trace the restransmit traffic that going
out of our cluster
35
Make sure with
your Cloud
Provider.

If this is can be
resolve. Latency
to that will DST
will much be
better.

36. How we utilise the tools so far – Trace our server that blocked from ISP
in indonesia by Goverment

36

37. How we utilise the tools so far – Get the Slow to Postgres queries

37

38. How we utilise the tools so far – Get the stat Redis queries

38

39. How we utilise the tools so far – Get the stat Redis queries

39

40. How we utilise the tools so far – Discovering something that we never
know it was exist

40
blanja.com is not
even exist
anymore.

41. Check your cluster!
41
kubectl gadget trace tcpdrop
kubectl trace dns|grep Non-Existen
kubectl trace tcpretrans

“It shouldn't hurt your cluster” i guess

42. Or try to visualize with Pixie
42
The red
dots that
mean
success
rates is
under~
50%

43. Or try to visualize with Pixie and get better latency each of your call
43
The red
line that
mean
success
rates is
~50%

44. Fun to watch eBPF Documentary
44

45. Any questions ?

You can find me at
●nicolas.my.id
●[email protected]
Thanks!
45

46. Talk #3

47. etcd and It’s super power
Irwan Shofwan

48. Agenda
●What is etcd?
●Superpower #1 Raft!
●Superpower #2 High Availability
●Superpower #3 Extensible
●What’s Next?

49. What is etcd?

50. What is etcd?
●etcd adalah key-value store
terdistribusi yang andal untuk
menyimpan data yang sangat
penting untuk suatu sistem
terdistribusi.
●etcd merupakan akronim dari etc
dan distributed
Source: https://github.com/etcd-io/etcd

51. What is etcd?
●etcd dikembangkan oleh Core OS untuk menangani concurrency control pada
upgrade sistem operasi
●etcd terinspirasi dari Apache Zookeeper dan Doozer
●etcd mulai digunakan kubernetes dari awal perkembangannya (v 0.4).
●etcd support digunakan dibanyak project lain seperti dapr, flannel, k3s, dll

52. What is etcd?
http://www.uml.org.cn/yunjisuan/202205174.asp

53. What is etcd?

54. Feature comparison
etcd Zookeeper Consul New Sql (Spanner,
CockroachDB, TiDB)
Concurrency Yes Not by default Yes Rare if any
Linearizable ReadsYes No Yes Sometimes
Multi-version
Concurrency
Control
Yes No No Sometimes
User permissionsRole Based ACL ACL Varies
Maximum reliable
database size
Several gigabytesHundreds of mb
(sometimes
several gb)
Hundreds of MBsTerabytes+

55. Superpower #1:
Raft

56. What is Raft?
●Raft merupakan suatu konsensus ataupun pendekatan yang digunakan oleh
etcd untuk memastikan data yang disimpan konsisten.
●Raft menggunakan pendekatan dengan memilih leader sebagai pengatur
replikasi dan request dari pengguna.
●Raft akan memastikan server lainnya (follower) menulis data yang sama
sebelum menulis miliknya sendiri

57. What is Raft?

58. What is Raft?

59. What is Raft?
In Search of an Understandable Consensus Algorithm (Extended Version), Diego Ongaro and John Ousterhout Stanford University

60. Why Raft?
●Raft memiliki performa
lebih baik daripada
paxos.
●Raft lebih mudah
dipahami.
https://www.semanticscholar.org/reader/31c45b6647429d5770093fc8cb90f18eafba2083
https://arxiv.org/abs/1911.01231

61. Why Raft?
●
https://www.semanticscholar.org/reader/31c45b6647429d5770093fc8cb90f18eafba2083

62. So, all distributed system use Raft then?
●Paxos
●Multi Paxos
●ZAB
●etc
https://www.researchgate.net/publication/224248727_Zab_High-performance_broadcast_for_primary-backup_systems
https://research.google.com/archive/chubby-osdi06.pdf

63. So, all distributed system use Raft then?
http://muratbuffalo.blogspot.com/2015/10/consensus-in-wild.html

64. Superpower #2:
High Availability

65. How etcd achieve High Availability?
●Raft!!!
●Node etcd akan melakukan leader election jika pada batas waktu tertentu
node follower tidak mendapatkan heartbeat dari leader.
●Node follower akan menjadi node candidate dan akan terpilih menjadi node
leader jika mendapatkan mayoritas vote

66. How etcd achieve High Availability?

67. Superpower #3:
Extensible

68. How etcd works?
KubeCon EU 2024: Unleash the Power of etcd: What Can an E[Xtensible]-etcd Bring? - Siyuan Zhang & Bogdan Kanivets

69. Etcd challenge
●Etcd dapat mengelola beberapa gigabyte
●Kebutuhan AI/ML membutuhkan data yang besar
KubeCon EU 2024: Unleash the Power of etcd: What Can an E[Xtensible]-etcd Bring? - Siyuan Zhang & Bogdan Kanivets

70. e-Etcd propose design
badger
sqlite
KubeCon EU 2024: Unleash the Power of etcd: What Can an E[Xtensible]-etcd Bring? - Siyuan Zhang & Bogdan Kanivets

71. What’s next?

72. What’s next?
●etcd sebagai opsi pilihan untuk pengembangan sistem terdistribusi
●Kustomisasi pada etcd sesuai kebutuhan dari use case kita
●Menggunakan beberapa bagian dari etcd untuk kebutuhan sistem distribusi
●Utilize other etcd superpower RBAC, MVCC, etc
Unleash the Power of etcd: What Can an E[Xtensible]-etcd Bring? - Siyuan Zhang & Bogdan Kanivets

73. Thank you

74. QnA

75. https://bit.ly/2024_June_Feedback_
CNCFID
We love to hear your
feedbacks and thoughts
for the event by fill out or
scan the feedback form!
Feedback Form

76. Photo
Session

77. Thank
You