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
クラウドネイティブな監視をMackerelで / Mackerel Day#2
Search
Sponsored
·
SiteGround - Reliable hosting with speed, security, and support you can count on.
→
FUJIWARA Shunichiro
December 23, 2019
Technology
4
5.9k
クラウドネイティブな監視をMackerelで / Mackerel Day#2
FUJIWARA Shunichiro
December 23, 2019
Tweet
Share
More Decks by FUJIWARA Shunichiro
See All by FUJIWARA Shunichiro
さくらのクラウドでのシークレット管理を考える/tamachi.sre#2
fujiwara3
1
240
Amazon ECS デプロイツール ecspresso の開発を支える「正しい抽象化」の探求 / YAPC::Fukuoka 2025
fujiwara3
13
9k
パフォーマンスチューニングのために普段からできること/Performance Tuning: Daily Practices
fujiwara3
8
6.2k
alecthomas/kong はいいぞ
fujiwara3
7
2.2k
ecspressoの設計思想に至る道 / sekkeinight2025
fujiwara3
12
3.4k
さくらのIaaS基盤のモニタリングとOpenTelemetry/OSC Hokkaido 2025
fujiwara3
3
2.8k
監視のこれまでとこれから/sakura monitoring seminar 2025
fujiwara3
11
5.7k
k6による負荷試験 入門から日常的な実践まで/Re:TechTalk #01
fujiwara3
2
480
困難を「一般解」で解く
fujiwara3
10
4.1k
Other Decks in Technology
See All in Technology
Agile Leadership Summit Keynote 2026
m_seki
1
640
CDKで始めるTypeScript開発のススメ
tsukuboshi
1
480
顧客の言葉を、そのまま信じない勇気
yamatai1212
1
360
茨城の思い出を振り返る ~CDKのセキュリティを添えて~ / 20260201 Mitsutoshi Matsuo
shift_evolve
PRO
1
330
15 years with Rails and DDD (AI Edition)
andrzejkrzywda
0
200
ClickHouseはどのように大規模データを活用したAIエージェントを全社展開しているのか
mikimatsumoto
0
260
顧客との商談議事録をみんなで読んで顧客解像度を上げよう
shibayu36
0
260
20260204_Midosuji_Tech
takuyay0ne
1
160
広告の効果検証を題材にした因果推論の精度検証について
zozotech
PRO
0
190
量子クラウドサービスの裏側 〜Deep Dive into OQTOPUS〜
oqtopus
0
130
登壇駆動学習のすすめ — CfPのネタの見つけ方と書くときに意識していること
bicstone
3
120
Claude_CodeでSEOを最適化する_AI_Ops_Community_Vol.2__マーケティングx_AIはここまで進化した.pdf
riku_423
2
600
Featured
See All Featured
brightonSEO & MeasureFest 2025 - Christian Goodrich - Winning strategies for Black Friday CRO & PPC
cargoodrich
3
100
Making the Leap to Tech Lead
cromwellryan
135
9.7k
The Invisible Side of Design
smashingmag
302
51k
StorybookのUI Testing Handbookを読んだ
zakiyama
31
6.6k
Cheating the UX When There Is Nothing More to Optimize - PixelPioneers
stephaniewalter
287
14k
The Limits of Empathy - UXLibs8
cassininazir
1
220
JavaScript: Past, Present, and Future - NDC Porto 2020
reverentgeek
52
5.8k
Mind Mapping
helmedeiros
PRO
0
88
The World Runs on Bad Software
bkeepers
PRO
72
12k
Understanding Cognitive Biases in Performance Measurement
bluesmoon
32
2.8k
Beyond borders and beyond the search box: How to win the global "messy middle" with AI-driven SEO
davidcarrasco
1
55
30 Presentation Tips
portentint
PRO
1
220
Transcript
ΫϥυωΠςΟϒͳࢹΛ Mackerel Ͱ 2019.12.23 Mackerel Day #2 @fujiwara
@fujiwara .BDLFSFMΞϯόαμʔ ʙ HJUIVCDPNLBZBDFDTQSFTTP "NB[PO&$4σϓϩΠπʔϧ HJUIVCDPNGVKJXBSBMBNCSPMM "84-BNCEBσϓϩΠπʔ ϧ
Game & Community
None
ΫϥυωΠςΟϒʁ $/$'$MPVE/BUJWF%FpOJUJPOW ΫϥυωΠςΟϒٕज़ɺύϒϦοΫΫϥυɺϓϥΠϕʔτΫϥυɺϋΠϒ ϦουΫϥυͳͲͷۙతͰμΠφϛοΫͳڥʹ͓͍ͯɺεέʔϥϒϧͳΞϓ ϦέʔγϣϯΛߏங͓Αͼ࣮ߦ͢ΔͨΊͷೳྗΛ৫ʹͨΒ͠·͢ɻ ͜ͷΞϓϩʔνͷදྫʹɺίϯςφɺαʔϏεϝογϡɺϚΠΫϩαʔϏεɺΠ ϛϡʔλϒϧΠϯϑϥετϥΫνϟɺ͓ΑͼએݴܕAPI͕͋Γ·͢ɻ IUUQTHJUIVCDPNDODGUPDCMPCNBTUFS%&'*/*5*0/NE
ίϯςφɺαʔϏεϝογϡɺϚΠΫϩαʔϏε… ͦΕΒͷٕज़Λ͍ͬͯΔ㱠ΫϥυωΠςΟϒ "Design for Failure" ͱͦΕΛ࣮ݱ͢ΔΈίϯϙʔωϯτɺͦΕΛ͍ ͜ͳ͢͜ͱ͕ͦ͜ΫϥυωΠςΟϒ োͷൃੜʹରͯࣗ͠ಈ෮چͰ͖ΔΑ͏ʹσβΠϯ͢ΔɻোͷൃੜʹΑΔϢʔ βʔӨڹ͕ͦͦͳ͍Α͏ʹΞʔΩςΫνϟΛσβΠϯ͢Δ2 IUUQTTQFBLFSEFDLDPNUPSJDMTEFTJHOGPSGBJMVSFJTUIFUSVFDMPVEOBUJWF
"Design for Failure" োΛఆͯ͠γεςϜΛσβΠϯ͢Δ w Πϯελϯεࢮ͵ w ϚωʔδυαʔϏεࢮ͵ ͚ͲେGBJMPWFS͢Δ w
σʔληϯλʔ͝ͱࢮ͵ كʹ͋Δ Ͳ͜·ͰΛఆ͠ɺͲ͔͜ΒఘΊΔ͔ ࢹͦΕΒΛંΓࠐΜͰઃܭ͢ΔˠΫϥυωΠςΟϒͳࢹ
ࢹରಈతʹ૿͑ͨΓݮͬͨΓ͢Δ ίϯςφσϓϩΠ͝ͱʹੜ·ΕมΘΔ ίϯςφͰͳͯ͘ w &$ͷΦʔτεέʔϦϯάͰ4QPUΠϯελϯεΛ͏ w ΘΓͱ͙͢མͪΔɺଞͷ্͕͕Δ ϚωʔδυαʔϏεෛՙʹԠͯ͡૿ݮͰ͖ΔΑ͏ʹͳ͖ͬͯͨ "VSPSB"VUP4DBMJOHͳͲ
Mackerel ͰΫϥυωΠςΟϒͷୈҰา ʮDPOOFDUJWJUZݕΛΊΔʯ
connec%vity ݕΛΊΔ ͜ͷΞϥʔτ ! $SJUJDBM͔͠ͳ͍ w ։ൃڥ͕ਂʹམͪͯඈͼى͖Δඞཁͳ͍ w ຊ൪ڥͰམͪͯαʔϏεʹӨڹ͢Δϗετͳ͍ Α͏ʹ࡞Δ
ϗετ͕ࢮ͵͜ͱΛલఏʹσβΠϯ͢ΔʹΫϥυωΠςΟϒ $SJUJDBMͳΞϥʔτʮαʔϏεͷܧଓੑʹӨڹ͢ΔͷʯͷΈ
Mackerel ͷΫϥυωΠςΟϒ͞ ! ϗετ͕૿͑ͯݮͬͯࣗಈͰैͰ͖Δ NBDLFSFMDPOUBJOFSBHFOU "84"[VSFΠϯςάϨʔγϣϯ ! ୀ͢ΔͱϗετϝτϦοΫ͕ݟ͑ͳ͘ͳΔ $16ͳͲҰ෦ͷϝτϦοΫͷΈΔɺΧελϜϝτϦοΫ ߹ܭฏۉͳͲظͰ͍͍͕ͨɺফ͑ͯ͠·͏
! ϗετ୯Ґ՝ۚ ϚΠΫϩϗετ ԁ݄ Ͱ424-BNCEBͷΑ͏ʹ ͍҆ɺϝτϦοΫ͕গͳ͍ରͷࢹʹͪΐͬͱߴ͍ʜ ૯ϝτϦοΫ՝ۚϓϥϯ͕΄͍͠
ϚωʔδυαʔϏεͷਐԽʹظͭͭ͠ ΫϥυωΠςΟϒͳࢹΛਐΊΔͨΊʹ ʮܺؒՈ۩044ʯ ϚωʔδυαʔϏεͷػೳαʔϏεؒ࿈ܞ͕ࣗͨͪͷӡ༻ʹ͓͍ͯෆेͳ ߹ʹɺͦͷ伱ؒΛຒΊͯΑΓΑ͍ӡ༻Λ࣮ݱ͢ΔͨΊʹ։ൃ͞ΕͨιϑτΣΞɻ ಛʹOSSͷͷΛࢦ͢ɻ3 IUUQTTQFBLFSEFDLDPNGVKJXBSBYJKJBOKJBKVPTTGBMTFTVTVNF IUUQTTQFBLFSEFDLDPNGVKJXBSBBXTEFWEBZUPLZP
ࠓհ͢Δ伱ؒՈ۩ OSS w NBQSPCF w NBDLFSFMQMVHJOQSPNFUIFVTRVFSZ IUUQTHJUIVCDPNGVKJXBSBNBDLFSFMQMVHJOQSPNFUIFVTRVFSZ IUUQTHJUIVCDPNGVKJXBSBNBQSPCF
ʲ՝ʳAWSΠϯςάϨʔγϣϯͰొ͞ΕͨϗετͰ mackerel-plugin ͰͷϝτϦοΫऔΓ͍ͨ
ྫɿAmazon RDS(MySQL)ʹରͯ͠ mackerel-plugin-mysql Λ࣮ߦ
ʲղ๏ʳͲ͔͜ͷϗετͷ mackerel-agent Ͱ plugin ࣮ߦʁ [plugin.metrics.rds01] command = "mackerel-plugin-mysql -host='rds01.***.ap-northeast-1.rds.amazonaws.com'
(ུ)" custom_identifier = "rds01.***.ap-northeast-1.rds.amazonaws.com" [plugin.metrics.rds02] command = "mackerel-plugin-mysql -host='rds02.***.ap-northeast-1.rds.amazonaws.com' (ུ)" custom_identifier = "rds02.***.ap-northeast-1.rds.amazonaws.com" ! ͜ͷϗετ͕མͪͨΒϝτϦοΫऩू͕ࢭ·Δ " ͋ͱ͔Β૿͑ͨϗετΛࢹ͢Δͷʹઃఆมߋ͕໘ # ࠷ۙNBDLFSFMBHFOU Λಈ͔͢ϗετ ͕ͳ͍͜ͱʜ ࢹରͷ૿ݮʹࣗಈै͍ͨ͠ʂ
maprobe w .BDLFSFMʹొ͞Εͨϗετʹରͯ͠ w ֎ܗࢹQJOHUDQIUUQ w NBDLFSFMQMVHJOΛ࣮ߦ ϗετϝτϦοΫͱͯ͠ߘ w ొࡁΈͷϗετϝτϦοΫΛू͠
αʔϏεϝτϦοΫͱͯ͠ߘ Λߦ͏ͨΊͷΤʔδΣϯτ
ʲղ๏ʳ maprobe Ͱ plugin ࣮ߦ probes: - service: production role:
RDS command: command: - 'mackerel-plugin-mysql' - '-host={{.Host.CustomIdentifier}}' - '-username=root' - '-password={{env "RDS_PASSWORD"}}' αʔϏεQSPEVDUJPO ϩʔϧ3%4ͷϗετશͯʹରͯ͠ NBDLFSFMQMVHJONZTRMΛ࣮ߦ ݁ՌΛݸʑͷϗετϝτϦοΫͱͯ͠.BDLFSFMૹ৴͢Δ
maprobe ରϗετͷ૿ݮʹࣗಈै ຖ.BDLFSFM"1*Λୟ͍ͯϗετΛݕࡧ ! ϗετͷ૿ݮʹࣗಈͰै %PDLFSίϯςφΞϦ㽂 docker pull fujiwara/maprobe 4ʹஔ͍ͨઃఆϑΝΠϧΛࣗಈͰ࠶ಡΈࠐΈ
! 4Λߋ৽͢ΕίϯςφϏϧυɾσϓϩΠෆ༻Ͱઃఆө maprobe agent --config s3://example.com/config.yaml IUUQTIVCEPDLFSDPNSGVKJXBSBNBQSPCF
ʲ՝ʳconnec%vity ΛΊͨΒ ϗετͷࢮ׆ࢹͲ͏͢Δʁ
ʲղ๏ʳmaprobe ͷϔϧενΣοΫػೳ ΈࠐΈͷϔϧενΣοΫػೳ QJOH 5$1 )551͕͋Δ probes: - service: production
role: EC2 ping: address: "{{ .Host.IPAddresses.eth0 }}" - service: production role: ElastiCacheRedis tcp: host: "{{ .Host.CustomIdentifier }}" port: 6379 send: "PING\n" expect_pattern: "PONG"
maprobe ͰͷϔϧενΣοΫ NBQSPCFͷϔϧενΣοΫ݁ՌϗετϝτϦοΫʹͳΔ DIFDLࢹͰͳ͍
check ࢹ͕Α͘ͳ͍ͱ͜Ζ(ࢲݟ) ઃఆมߋ͕ϑΝΠϧͷमਖ਼ σϓϩΠ ʮͪΐͬͱ͍·͚ͩࢹP⒎ᮢมߋʯ͕͍͠ ᮢͷධՁํ๏͕ϓϥάΠϯ͝ͱʹ·ͪ·ͪ --critical-underʮҎԼʯ͔ʮະຬʯ͔ʜ ҰʹଟͷϗετͰൃใ͕ͪ͠ ेϗετ͔ΒDIFDLࢹࣦഊ͕དྷͯݪҼݸͩͬͨΓ OUQͷ࣌ࠁͣΕɺEBFNPOͷઃఆ
EFQMPZ ϛεʜ
check ࢹ = metric ࢹͷಛघͳύλʔϯ ϝτϦοΫΛอଘͯͦ͠ΕΛධՁ͢Εಉ͜͡ͱ͕Ͱ͖Δ ϝτϦοΫࢹɺࣜࢹΛ׆༻͢Δ
ྫɿ ping ʹΑΔࢮ׆ࢹ sum(role(production:EC2, ping.count.failure)) QSPDVUDJPO&$ͷ͍ͣΕ͔ͷϗετʹQJOH͕ࣦഊͨ͠ΒXBSO Կ͔མͪͯαʔϏε͕ఏڙͰ͖͍ͯΕ$SJUJDBMͰͳ͍
ྫɿ job queue ͷཹ job ΛΞϥʔτ sum(role(production:job-queue, custom.gearmand.queue.*.total)) ෳͷϗετʹKPCRVFVF͕͋Δ ཹKPCΛϝτϦοΫʹ͍ͯ͠Δ
ཷ·Δͱ͖શͯͷRVFVF͕ཷ·Δ͜ͱ͕ଟ͍ DIFDLࢹͰݸผʹΞϥʔτ͢Δͱશ෦ͷϗετͰൃใ͕ͪ͠ ߹ܭΛݟΔ͜ͱͰશମͷॲཧঢ়گΛѲ͢Δ
ʲ՝ʳୀ͢Δͱফ͑ͯ͠·͏ ϗετϝτϦοΫΛ͍͍ͨ
ΧελϜϝτϦοΫϗετ͕ୀ͢Δͱফ͑ͯ͠·͏
ʲղ๏ʳmaprobe ͰϗετϝτϦοΫΛूอଘ aggregates: - service: production role: push-server metrics: -
name: custom.push.messages.sent outputs: - func: sum name: custom.push.messages.total_sent αʔϏεQSPEVDUJPO ϩʔϧQVTITFSWFSʹରͯ͠ ϗετϝτϦοΫͷQVTINFTTBHFTTFOUΛશऔಘ ˠԋࢉͨ݁͠ՌΛαʔϏεϝτϦοΫͱͯ͠อଘ͢Δ
maprobe aggregate func/ons ݱࡏTVN NJO NBY BWFSBHF DPVOUΛαϙʔτ QFSDFOUJMF͋ͬͨ΄͏͕Αͦ͞͏͚ͩͲະ࣮ ΧελϜϝτϦοΫ͕ফ͑ͳ͚ΕࣜάϥϑͰ݁ͳͷͰԿଔ
ʲ՝ʳͬͱΫϥυωΠςΟϒͳ Ϧιʔεͷࢹ
ϚΠΫϩαʔϏεʂ αʔϏεϝογϡʂ &OWPZΛ͍࢝Ί͍ͯΔͷͰɺϝτϦΫεΛऔΓ͍ͨ ͱ͋Δ&OWPZͷ/statsΛୟ͘ͱʜ $ curl -s x.x.x.x:9901/stats ... cluster.web.default.total_match_count:
1 cluster.web.external.upstream_rq_200: 988 cluster.web.external.upstream_rq_2xx: 988 cluster.web.external.upstream_rq_302: 13 cluster.web.external.upstream_rq_3xx: 13 cluster.web.external.upstream_rq_400: 3 cluster.web.external.upstream_rq_403: 26 cluster.web.external.upstream_rq_404: 5 cluster.web.external.upstream_rq_4xx: 34 .... IUUQTFOWPZQSPYZJP
͜ΕΛશ෦ Mackerel ʹૹΕΑ͍ʁ cluster.web.default.total_match_count: 1 cluster.web.external.upstream_rq_200: 988 cluster.web.external.upstream_rq_2xx: 988 cluster.web.external.upstream_rq_302:
13 cluster.web.external.upstream_rq_3xx: 13 cluster.web.external.upstream_rq_400: 3 cluster.web.external.upstream_rq_403: 26 cluster.web.external.upstream_rq_404: 5 cluster.web.external.upstream_rq_4xx: 34 ...
͍߹Θͤ ϑΟʔυόοΫϑΥʔϜ͔Βؾܰʹฉ͍ͯΈͨ ۙʑenvoy(https://www.envoyproxy.io/)Λಋೖ༧ఆͳͷͰ͕͢ɺMackerel ެࣜͱͯ͠envoy statsऔಘϓϥάΠϯΛެ։͞ΕΔ༧ఆ͋Γ·͢Ͱ͠ΐ͏͔ʁ envoyͷϓϥάΠϯͰ͕͢ɺݱ࣌Ͱެࣜͱͯ͠ެ։͢Δ༧ఆ͍͟͝ ·ͤΜɻ ! ࡞Δ͔ʜʜ
͔͠͠ Envoy େྔͷϝτϦοΫΛు͖ग़͢ $ curl -s x.x.x.x:9901/stats | wc -l
337 ͜ΕΛશ෦.BDLFSFMʹ͍࣋ͬͯ͘ͱʜ ϚΠΫϩϗετϝτϦοΫϗετˠϗετ૬ &OWPZ͕͍Δϗετ͝ͱʹYԁ ੫ผ
࡞ઓมߋ શ෦Λ͍࣋ͬͯ͘ͷίετ͕ݫ͍͠ ͱ͍͑ͲΜͳ͕΄͍͔͠Α͔͍ͬͯ͘ͳ͍ &OWPZӡ༻ܦݧ͕ઙ͍ͷͰɺӡ༻͠ͳ͕ΒݟΔΛܾΊ͍ͨ QMVHJOΛ࡞ͬͯ༗༻ͳϝτϦοΫΛऔΔʹӡ༻ܦݧ͕ඞཁ ͭ·ΓͱΓ͋͑ͣશ෦औΓ͍ͨ
Prometheus Ͳ͏͔ʁ 1VMMܕϝτϦοΫऩूɾࢹπʔϧ อଘͨ͠Λ1SPN2-ΫΤϦͰॊೈʹՃͯ͠औಘͰ͖Δ ظؒͷϝτϦοΫอଘ͋·Γߟྀ͞Ε͍ͯͳ͍ (SBGBOBͱ͔ͰՄࢹԽ͕ී௨ʁ
! Envoy → Prometheus → Mackerel զʑ.BDLFSFMͰΞϥʔτ͍ͨ͠͠άϥϑݟ͍ͨ ͲͷΛऔΔ͖͔͕ݟ͑ͳ͍ͷͰɺۙશ෦औ͓͖͍ͬͯͨ &OWPZˠ1SPNFUIFVT
͜Ε͋Δ 1SPNFUIFVTˠΫΤϦ݁ՌΛNFUSJDQMVHJOܗࣜͰग़ྗ NFUSJDQMVHJOܗࣜͷग़ྗΛ.BDLFSFMʹอଘ ͜Ε͋Δ ͜ͷ͚ͩ࡞ΕΑ͍ͷͰʂ Γͳ͍1SPNFUIFVTʹΫΤϦͯ͠औΕΑ͍
mackrel-plugin-prometheus-query ࡞Γ·ͨ͠ HJUIVCDPNGVKJXBSBNBDLFSFMQMVHJOQSPNFUIFVTRVFSZ $ mackerel-plugin-prometheus-query \ -query "up" \ -metric-key-format
"promq.{job}.{instance}" promq.web.10_1_129_175_9901 1 1575941187 promq.web.10_1_130_170_9901 1 1575941187 promq.web.10_1_131_53_9901 1 1575941187 promq.prometheus.localhost_9090 1 1575941187
Prometheus ͷΫΤϦྫ ྫɿؒͷVQTUSFBNͷϦτϥΠճΛٻΊΔΫΤϦ sum( delta( envoy_cluster_upstream_rq_retry{envoy_cluster_name="web"}[1m] ) ) .BDLFSFMʹϗετ୯ҐͰͳ͘શମͷΛૹΔ ݸʑͷ1SPNFUIFVTΛݟΕ͋ΔͷͰ
plugin ͷग़ྗΛ mkr throw Ͱ͛Δ $ mackerel-plugin-prometheus-query \ -query 'sum(delta(envoy_cluster_upstream_rq_retry_success
{envoy_cluster_name="web"}[1m]))' \ -metric-key-format "envoy.web.upstream.retry.success" \ | mkr throw --service production
·ͱΊ ΫϥυωΠςΟϒͳࢹͱ োΛڐ༰͢Δઃܭશମͷ݈શੑΛݟΔ ͦͷಓ۩ͱͯ͠ͷܺؒՈ۩044 w NBQSPCF ࣗಈैɺ֎ܗࢹɺQMVHJOࢹɺϝτϦοΫू w NBDLFSFMQMVHJOQSPNFUIFVTRVFSZ TIPSUUFSN1SPNFUIFVTͰ
MPOHUFSN BMFSUJOH.BDLFSFMͰ