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
coil.pdf
Search
Sponsored
·
Your Podcast. Everywhere. Effortlessly.
Share. Educate. Inspire. Entertain. You do you. We'll handle the rest.
→
Akihiro Ikezoe
February 19, 2019
Technology
3.2k
5
Share
Embed
Copy iframe code
Copy JS code
Copy link
Start on current slide
coil.pdf
Akihiro Ikezoe
February 19, 2019
More Decks by Akihiro Ikezoe
See All by Akihiro Ikezoe
Kubernetesコントローラーのパフォーマンスチューニング
zoetrope
4
2.3k
Kubernetes Admission Webhook Deep Dive
zoetrope
8
1.6k
Kubernetesオペレータのアンチパターン&ベストプラクティス
zoetrope
11
5k
Production-Ready Kubernetesに至るまでの3年間とこれから
zoetrope
4
960
オンプレKubernetesでMySQLクラスタの運用を自動化するためにOperatorを自作している話
zoetrope
5
2.5k
サイボウズを支える技術~インフラ刷新プロジェクトNecoを中心に紹介~
zoetrope
1
1.3k
Kuebernetesクラスタのマルチテナンシーベストプラクティス
zoetrope
8
6.9k
クラウドネイティブなチームづくり
zoetrope
7
4k
Open Policy Agent / Gatekeeper 勉強会
zoetrope
5
3k
Other Decks in Technology
See All in Technology
IHV like なユースケースへのOpenID Connect 関連仕様の適用事例
optim
0
350
ハッカソンで入賞した話 @ Findy LT
asari194617
0
1.5k
Kiro5兄弟のいまどきのセキュリティ基礎知識
kentapapa
0
370
AWSとGitHub Actionsの責任境界と 組織で安全に使用する取り組み
nealle
0
210
RapidCopy2 Matrix I/Oエンジンによるファイルコピーソフトウェアの設計と実装
kengosawa2
1
470
GopherCon @シアトル に行ってきました
logica0419
0
330
When Token Pruning is Worse than Random: Understanding Visual Token Information in VLLMs
sansantech
PRO
0
130
研究開発部の紹介 / Sansan R&D Profile
sansan33
PRO
4
24k
Windows の互換機能 - 古いプログラムはなぜ動作できるのか
murachiakira
PRO
0
120
最新技術に積極チャレンジ!EKS共通基盤のこれまでとこれから
daitak
0
250
「ミスを許さない手順書」を作ってみた 〜 個人的にはこれ以上できることはあまりなさそう/20260827-ssmjp-operation-procedure-update
opelab
12
8.9k
フィジカルAIの知識ゼロの僕でも AIを使えばここまでできた
tatsuya1970
0
200
Featured
See All Featured
The #1 spot is gone: here's how to win anyway
tamaranovitovic
3
1.1k
<Decoding/> the Language of Devs - We Love SEO 2024
nikkihalliwell
1
300
The Myth of the Modular Monolith - Day 2 Keynote - Rails World 2024
eileencodes
28
3.6k
Music & Morning Musume
bryan
47
7.3k
The Mindset for Success: Future Career Progression
greggifford
PRO
0
480
Lessons Learnt from Crawling 1000+ Websites
charlesmeaden
PRO
1
1.5k
Designing for Timeless Needs
cassininazir
1
460
AI Search: Implications for SEO and How to Move Forward - #ShenzhenSEOConference
aleyda
1
1.3k
Navigating the Design Leadership Dip - Product Design Week Design Leaders+ Conference 2024
apolaine
2
410
YesSQL, Process and Tooling at Scale
rocio
174
15k
Visual Storytelling: How to be a Superhuman Communicator
reverentgeek
2
630
So, you think you're a good person
axbom
PRO
2
2.1k
Transcript
େنKubernetesΫϥελ͚ʹ CNIϓϥάΠϯΛࣗ࡞ͨ͠ αΠϘζגࣜձࣾ ఴ ໌ 1 Kubernetes Meetup Tokyo #16
2019/02/19
ຊͷൃද༰ • Πϯϑϥ৽ϓϩδΣΫτ Neco ͷհ • ͳͥCNIϓϥάΠϯΛࣗ࡞ͨ͠ͷ͔ʁ • ࣗ࡞CNIϓϥάΠϯ coil
ͷհ • CNIϓϥάΠϯ։ൃͰಘΒΕͨݟ • ·ͱΊ 2
Πϯϑϥ৽ϓϩδΣΫτ Neco 3
Πϯϑϥ৽ϓϩδΣΫτNecoͱ • Kubernetes Λಋೖͯ͠ ͷΠϯϑϥ Λ৽͢ΔϓϩδΣΫτ • ͱ •
kintoneGaroon, OfficeͳͲͷαʔϏεΛSaaSͱͯ͠ఏڙ • ಋೖاۀ2.5ສࣾɺϢʔβʔ100ສਓ • 2011ϦϦʔε • VMϕʔεͷΞʔΩςΫνϟ • σʔληϯλʔΛआΓͯɺ1,000Ҏ্ͷαʔόʔΛࣗલͰӡ༻ 4
Πϯϑϥ৽ϓϩδΣΫτNecoͱ • త • ϝϯςφϯείετͷେ෯ݮ (NoOpsΛࢦ͢) • εέʔϥϏϦςΟͷ্ • αʔόʔͷूੑ্
• ΞϓϦέʔγϣϯ։ൃνʔϜ͕σϓϩΠɾӡ༻ʹࢀՃ • ΄ͱΜͲͷՌΛOSSͱͯ͠ެ։ 5
NecoͷΞʔΩςΫνϟ Kubernetes CoreOS Node LB Prometheus squid CoreOS Node CoreOS
Node CoreOS Node CoreOS Node Boot Server CKE sabakan CoreDNS 1,000نͷαʔόʔ ઍʙͷΞϓϦέʔγϣϯίϯςφ 5ͷϒʔταʔόʔ Kubernetesͷ σϓϩΠ ཧ ܧଓతσϦόϦʔ app app MySQL Elastic search 6 CoreOS Node neco updater Ubuntu Argo CD Rook
NecoΛࢧ͑ΔιϑτΣΞͨͪ • sabakan • αʔόʔػࡐͷϥΠϑαΠΫϧཧͱϓϩϏδϣχϯάͷࣗಈԽΛ͓͜ͳ͏ɻ • BIOSͷઃఆɺOSͷωοτϒʔτɺσΟεΫ҉߸Խ֤छγεςϜιϑτΣΞͷ ηοτΞοϓΛࣗಈతʹ͓͜ͳ͏ɻ • CKE
(Cybozu Container Engine) • KubernetesΫϥελͷࣗಈߏஙɾӡ༻Λ͓͜ͳ͏πʔϧɻ • sabakan͕ߏஙͨ͠αʔόʔʹKubernetesΛࣗಈతʹσϓϩΠ͢Δɻ • ΤϥʔΛࣗಈम෮ͨ͠ΓɺނোػࡐΛΫϥελ͔ΒऔΓ֎͢ͳͲͷӡ༻ΛࣗಈԽɻ • neco-updater • ΠϯϑϥͷܧଓతσϦόϦʔπʔϧɻ • GitHubͷϦϦʔεใΛνΣοΫ͠ɺCKEsabakanΛ͡Ίͱ͢Δ֤छιϑτ ΣΞͷσϓϩΠɺCoreOSΠϝʔδͷߋ৽ͳͲΛࣗಈతʹ࣮ࢪ͢Δɻ 7
ͳͥCNIϓϥάΠϯΛ ࣗ࡞ͨ͠ͷ͔ʁ 8
ͳͥCNIϓϥάΠϯΛࣗ࡞ͨ͠ͷ͔ʁ • NecoϓϩδΣΫτͷωοτϫʔΫߏʹϚονͨ͠ωοτϫʔΫϓ ϥάΠϯ͕ඞཁ • طଘͷϓϥάΠϯΛબఆ͕ͨ͠ɺશʹཁ݅ʹϚον͢Δͷ͕ͳ ͔ͬͨɻ • CNIϓϥάΠϯෳΛΈ߹Θͤͯར༻͢Δ͜ͱ͕Մೳɻඞཁͳ෦ ͚ͩࣗ࡞͠Α͏ɻ
9
NecoͷωοτϫʔΫߏ Rack1 Rack2 Rack3 • CLOSΞʔΩςΫνϟ • ϑϥοτͳL3ωοτϫʔΫ • ֤ϊʔυʹಉҰϗοϓͰ౸ୡՄೳ
• East-WestτϥϑΟοΫͷ૿େʹର͠ ͯεέʔϧՄೳ • BGPʹΑΔϧʔςΟϯά • AS per Rack • ECMPʹΑΔܦ࿏Խ • BFDʹΑΔߴͳܦ࿏ऩଋ • ৄ͘͠ϒϩάͰ • https://blog.cybozu.io/entry/2018/1 1/01/113000 10
CNIϓϥάΠϯͷબఆ • σʔληϯλʔωοτϫʔΫͱ߹ΘͤͯɺKubernetesͷωοτϫʔΫʹBGP Λ࠾༻ͯ͠ޮతʹϧʔςΟϯάΛ͓͜ͳ͍͍ͨɻ • Calico • ։ൃ͕׆ൃͰػೳ๛ɻ࠾༻࣮ଟ͍ɻ • BGP
SpeakerΛแ͍ͯ͠Δ͜ͱɺେنΫϥελͰܦ࿏͕૿େ͢Δ ͜ͱͳͲ͕ݒ೦ɻ • Romana • ػೳతʹཁ݅Λຬ͍ͨͯ͠Δɻ • etcd v3ະରԠɻ։ൃ͕׆ൃͰͳ͘࠷৽ͷKubernetesʹैͰ͖͍ͯͳ͍ɻ 11
ϓϥάΠϯͷΈ߹Θͤ • ίϯςφωοτϫʔΫͰղܾ͖͢՝͕ଟ͋Δɻ • IPΞυϨεཧ (IPAM) • ωοτϫʔΫଓੑ • ωοτϫʔΫϙϦγʔ
• ଳҬ੍ݶ • ͯ͢ͷΛ1ͭͷϓϥάΠϯͰղܾ͢Δඞཁͳ͍ɻ • ෳͷϓϥάΠϯΛΈ߹ΘͤΔྫ: Canal • IPAMCNIϓϥάΠϯͷඪ४ػೳΛར༻ • ωοτϫʔΫଓੑ flannel Λར༻ • ωοτϫʔΫϙϦγʔ Calico Λར༻ 12
Ͳ͜Λࣗ࡞͢Δඞཁ͕͋Δͷ͔ʁ • ωοτϫʔΫϙϦγʔ • ࣮͕େมɻಠࣗੑΛग़͢ඞཁͳ͍ɻ CalicoCiliumͳͲΛར༻ • IPΞυϨεཧ (IPAM)
• ϧʔςΟϯάςʔϒϧͷ૿େΛආ͚ΔͨΊͷΈ͕ඞཁɻ • PodʹάϩʔόϧIPΞυϨεΛׂΓͯΔػೳཉ͍͠ɻ ࣗ࡞ • ωοτϫʔΫଓ • ϊʔυͷ֎ͱͷܦ࿏ަɺDCͷϧʔςΟϯάιϑτΣΞʹͤΔ ϊʔυͷϧʔςΟϯά෦Λࣗ࡞ 13
ࣗ࡞CNIϓϥάΠϯ coil 14
coilͱ • αΠϘζ͕։ൃ͢ΔCNIϓϥάΠϯ • OSSͱͯ͠ެ։ • https://github.com/cybozu-go/coil • ಛ •
CNI Spec v0.3.1 (Kubernetes v1.13)ʹରԠ • GoݴޠͰ࣮ɺόοΫΤϯυʹ etcd Λ࠾༻ • Linux only, Kubernetes only, IPv4 only • IPAMͱϊʔυͷϧʔςΟϯάػೳͷΈΛఏڙ • ϧʔςΟϯάιϑτΣΞΛแ͠ͳ͍ • େنΫϥελͰͷར༻Λߟྀ 15
େنΫϥελͰར༻͢ΔͨΊʹ • ΦʔόϨΠωοτϫʔΫར༻͠ͳ͍ • VXLANͳͲͷΦʔόʔϨΠωοτϫʔΫεϧʔϓοτ͕མͪΔɻ ୯७ͳϧʔςΟϯάʹΑΓωοτϫʔΫͷଓੑΛཱ֬ • Linux BridgeΛར༻͠ͳ͍
• Linux BridgeΛ͏ͱCPU༻͕૿͑ͯ͠·͏ɻ ϧʔςΟϯάςʔϒϧʹvethͷϧʔτΛՃ͢Δɻ • etcd API v3ͷΈʹରԠ • etcd API v2ɺଟͷΫϥΠΞϯτ͔Βଓ͞Εͨ࣌ʹύϑΥʔϚϯε͕མͪΔɻ • ܦ࿏ͷ૿େΛ͙ • Pod͝ͱʹܦ࿏Λࠂ͢Δͱܦ࿏͕૿େͯ͠͠·͏ɻ ΞυϨεϒϩοΫͱ͍͏ΈʹΑΓɺαϒωοτ͝ͱͷܦ࿏Λࠂɻ 16
Node coilͷߏཁૉ • etcd • ׂΓͯͨΞυϨεใΛཧ • coil-controller • k8s্ͷDeployment
• ΘΕͳ͘ͳͬͨΞυϨεϒϩοΫͷղ์ • coil: • CNIϓϥάΠϯຊମ • PodͷvethͷՃɾআɺIPΞυϨεׂΓͯɺ ϧʔςΟϯάઃఆͳͲΛ࣮ࢪ • coil-node: k8s্ͷDaemonSet • coild: • ϊʔυ͝ͱͷΞυϨεཧϧʔςΟϯάͷઃఆΛ࣮ࢪ • coil-installer: • coilઃఆϑΝΠϧͷΠϯετʔϧΛ࣮ࢪ 17 <DeamonSet> coil-node <CNI> coil etcd House Keeping <Deployment> coil-controller coild coil installer conf ηοτΞοϓ
ΞυϨεϒϩοΫ (Inspired by Romana) • coilͰΞυϨεϒϩοΫͱ͍͏ΈΛ औΓೖΕɺαϒωοτ୯Ґ(ྫ: /28)Ͱܦ࿏ Λࠂ͢Δ͜ͱͰϧʔςΟϯάςʔϒϧͷ ංେԽΛճආ͍ͯ͠Δɻ
18 Node2 Pod 10.0.1.16/32 Node1 Pod 10.0.1.0/32 Pod 10.0.1.17/32 Pod 10.0.1.1/32 BGP Router 10.0.1.0/28 -> Node1 10.0.1.16/28 -> Node2 αϒωοτ୯ҐͰ ܦ࿏Λࠂ PodͷΞυϨεϨϯδ: 10.0.1.0/24 ΞυϨεϒϩοΫ: 10.0.1.0/28 ΞυϨεϒϩοΫ: 10.0.1.16/28 ΞυϨεϒϩοΫ: 10.0.1.32/28 ΞυϨεϒϩοΫ: 10.0.1.48/28 ΞυϨεϒϩοΫ: 10.0.1.64/28 ΞυϨεϒϩοΫ 10.0.1.0/28 ΞυϨεϒϩοΫ 10.0.1.16/28 ɿ ϊʔυ͝ͱʹ ΞυϨεϒϩοΫ ΛׂΓͯ ׂΓͯՄೳͳ ΞυϨεΛ ϒϩοΫͱ͍͏ ୯Ґʹׂ PodͷΞυϨε ΞυϨεϒ ϩοΫ͔Βׂ Γͯ
coilͷॲཧͷྲྀΕ 1. kubelet ͕ࢦࣔΛड͚ͯ Pod Λ࡞ 2. CNIϓϥάΠϯ coil Λ࣮ߦ
3. coil ͔Β coild ʹIPΞυϨεΛཁٻ 4. etcd ͔ΒΞυϨεϒϩοΫΛׂΓͯΔ 5. ܦ࿏ΛϧʔςΟϯάςʔϒϧʹॻ͖ग़͢ 6. ϧʔςΟάςʔϒϧΛಡΈࠐΈܦ࿏Λࠂ 7. coild ͔Β coil ʹIPΞυϨεΛฦ͢ 8. Pod ͷ netns ʹ veth ͷϖΞΛ࡞͠ɺIPΞυ ϨεͷׂΓͯͱϧʔτͷઃఆΛ͓͜ͳ͏ Node <DeamonSet> coild <CNI> coil kubelet Pod etcd routing table BGP Speaker eth0 veth BGP Router ᶃ ᶃ ᶄ ᶅ ᶆ ᶇ ᶈ ᶈ ᶊ ᶉ 19
boot-taint • CNIϓϥάΠϯ͕ηοτΞοϓ͞Ε͍ͯͳ͍ϊʔυ ʹPod͕εέδϡʔϧ͞ΕΔͱࠔΔɻ • kubeletͷىಈΦϓγϣϯʹ —register-with-taints Λ ࢦఆͯ͠ɺىಈޙͷϊʔυʹPodΛεέδϡʔ ϧͰ͖ͳ͍Α͏ʹ͓ͯ͘͠ɻ
• coil ͷηοτΞοϓ͕ྃͨ͠Β taints Λআ͠ɺ PodΛεέδϡʔϦϯάՄೳʹ͢Δɻ 20 taints/tolerations Kubernetesͷػೳɻ ϊʔυʹtaintsΛ༩͢ Δ͜ͱͰɺPodͷεέ δϡʔϦϯά࣮ߦΛ ېࢭ͢Δ͜ͱ͕Ͱ͖Δɻ ಛఆͷtolerations͕ ༩͞ΕͨPodͷΈεέ δϡʔϦϯάՄೳͱͳ Δɻ
CNIϓϥάΠϯ։ൃͰ ಘΒΕͨݟ 21
։ൃσόοά͕େมͰʁ • NecoϓϩδΣΫτͰɺσʔληϯλʔͷωοτϫʔΫߏΛؙ͝ͱιϑτ ΣΞͰԾԽͨ͠ڥΛ༻ҙ͓ͯ͠ΓɺखݩͰ؆୯ʹಈ࡞֬ೝΛ͓͜ͳ͏͜ͱ ͕Ͱ͖Δɻ • γϯϓϧͳL3ωοτϫʔΫͳͷͰௐ͍ࠪ͢͠ɻ nsenterͱtcpdump͕͋ΕɺେͷௐࠪͰ͖Δɻ 22
KubernetesͷใΛಘΔʹʁ • CNIϓϥάΠϯKubernetesઐ༻ͷͷͰͳ͍ͨΊɺPodͷใΛऔಘ͢Δ ༷ఆΊΒΕ͍ͯͳ͍ɻ • https://github.com/containernetworking/cni/issues/606 • ݱঢ়Kubernetes͔ΒCNIϓϥάΠϯΛݺͼग़͢ࡍʹɺCNI_ARGS Ͱ K8S_POD_NAME
K8S_POD_NAMESPACE ͳͲͷใΛ͍ͯ͠Δɻ • ໌֬ʹ༷ͱͯ͠ఆ·͍ͬͯΔΘ͚Ͱͳ͍ͷͰ࣮ΛಡΈղ͘ඞཁ͕͋ͬͨɻ (dockershimcontainerdͳͲɺίϯςφϥϯλΠϜ͝ͱʹͦΕͧΕ࣮ͯ͠Δ) 23
CNIϓϥάΠϯ͕ಡΈࠐ·Εͳ͍ • Kubernetes/etc/cni/net.dʹ͓͍ͯ͋ΔઃఆΛݩʹϓϥάΠϯΛ࣮ߦ͢Δɻ • ઃఆϑΝΠϧͷߋ৽ޙʹΞϓϦέʔγϣϯͷPodΛ࡞ͨ͠ͱ͜Ζɺઃఆͨ͠ ϓϥάΠϯ͕ར༻͞Εͳ͍͜ͱ͕͋Δɻ • kubeletͰɺ5ඵ͝ͱʹίϯςφϥϯλΠϜͷεςʔλεߋ৽Λߦͳ͍ͬͯΔɻ ͜ͷͱ͖ʹCNIϓϥάΠϯಡΈࠐΜͰ͍Δɻ 24
Kubernetes 1.13ͰPodؒ௨৴͕Ͱ͖ͳ͍ • KubernetesΛv1.13ʹΞοϓάϨʔυͨ͠ΒɺPodؒͷ௨৴͕Ͱ͖ͳ͘ͳͬͨɻ • kube-proxyͷIPVSϞʔυͷ࣮ͰɺLinuxͷΧʔωϧύϥϝʔλ(sysctl)ͷઃఆ͕ มߋ͞Ε͍ͯͨ (ϦϦʔεϊʔτʹهड़ͳ͠) • net.ipv4.conf.all.arp_ignore:
0 -> 1 • net.ipv4.conf.all.arp_announce: 0 -> 2 • ͍͔ͭ͘ͷCNIϓϥάΠϯʹӨڹ͕ग़͍ͯΔ • https://github.com/kubernetes/kubernetes/issues/71555 25
ͱղܾํ๏ Node Pod ens3 eth0 veth 192.168.1.11 IP: 10.0.1.1/32 GW:
192.168.1.11 ARP: 192.168.1.11 Node Pod ens3 eth0 veth 192.168.1.11 IP: 10.0.1.1/32 GW: 169.254.1.1 169.254.1.1/32 arp_ignore=0 veth͕ens3ͷMACΞυϨεΛฦ͢ arp_ignore=1 veth͕ens3ͷMACΞυϨεΛฦ͞ͳ͍ Pod͔ΒNodeʹͭͳ͕Βͳ͍ʂ vethʹϦϯΫϩʔΧϧΞυϨεΛׂΓ ͯɺPodͷσϑΥϧτήʔτΣΠ Λ͜ͷΞυϨεʹมߋͨ͠ɻ मਖ਼લ मਖ਼ޙ 26
·ͱΊͱࠓޙ 27
·ͱΊͱࠓޙ • ·ͱΊ • େنͳKubernetesΫϥελͰͷར༻Λߟྀͨ͠CLOSΞʔΩςΫνϟ͚ CNIϓϥάΠϯ coil Λ։ൃͨ͠ • coil
ͷ࣮γϯϓϧͰಡΈ͍͢ͱࢥ͏ͷͰɺCNIϓϥάΠϯΛֶͼ͍ͨਓ ʹΦεεϝ • ࠓޙ • େنΫϥελʹద༻ͯ҆͠ఆੑΛݕূ • νϡʔτϦΞϧͳͲΛ༻ҙͯ͠ར༻͘͢͠ 28
We are hiring! • NecoϓϩδΣΫτͷ࠾༻ใ • https://cybozu.co.jp/company/job/recruitment/list/neco_project.html • εΩϧνΣοΫγʔτ •
https://gist.github.com/ymmt2005/bd92296166e52d1beba9df8ac516a9db • NecoϓϩδΣΫτͰʹ͚ͭΒΕΔεΩϧΛհ • ଟ༷ͳಇ͖ํ • શϦϞʔτϫʔΫɺि20࣌ؒۈɺଞࣾͱͷ݉ۀͳͲɺ͍Ζ͍Ζͳಇ͖ํ Λ͍ͯ͠Δϝϯόʔ͕ॴଐ 29