S\u00fcrekli geli\u015fen web uygulamalar\u0131 d\u00fcnyas\u0131nda, kullan\u0131c\u0131 deneyimini kesintiye u\u011fratmadan yeni \u00f6zellikler sunmak veya hata d\u00fczeltmeleri yapmak, modern yaz\u0131l\u0131m geli\u015ftirme ekipleri i\u00e7in kritik bir gerekliliktir. Geleneksel da\u011f\u0131t\u0131m y\u00f6ntemleri genellikle uygulaman\u0131n k\u0131sa bir s\u00fcreli\u011fine eri\u015filemez olmas\u0131na neden olurken, “s\u0131f\u0131r kesinti da\u011f\u0131t\u0131m\u0131” (zero-downtime deployment) bu sorunu ortadan kald\u0131rmay\u0131 hedefler. \u00d6zellikle Ruby on Rails gibi dinamik ve veritaban\u0131 odakl\u0131 uygulamalar i\u00e7in s\u0131f\u0131r kesinti da\u011f\u0131t\u0131m\u0131, i\u015f s\u00fcreklili\u011fi ve kullan\u0131c\u0131 memnuniyeti a\u00e7\u0131s\u0131ndan vazge\u00e7ilmezdir. Bu makalede, pop\u00fcler bir Rails uygulama sunucusu olan Puma ve s\u00fcre\u00e7 y\u00f6neticisi Foreman’\u0131 kullanarak Rails uygulamalar\u0131n\u0131z i\u00e7in s\u0131f\u0131r kesinti da\u011f\u0131t\u0131mlar\u0131n\u0131 nas\u0131l yap\u0131land\u0131raca\u011f\u0131n\u0131z\u0131 ayr\u0131nt\u0131l\u0131 bir \u015fekilde inceleyece\u011fiz.<\/p>\n
G\u00fcn\u00fcm\u00fcz\u00fcn rekabet\u00e7i dijital pazar\u0131nda, bir web uygulamas\u0131n\u0131n birka\u00e7 saniye bile eri\u015filemez olmas\u0131 ciddi sonu\u00e7lar do\u011furabilir. Kullan\u0131c\u0131lar, h\u0131zl\u0131 ve kesintisiz hizmet beklerler. Bu beklentiyi kar\u015f\u0131lamak, sadece iyi bir kullan\u0131c\u0131 deneyimi sunmakla kalmaz, ayn\u0131 zamanda i\u015f hedeflerine ula\u015fmak i\u00e7in de temel bir gerekliliktir.<\/p>\n
Bir uygulaman\u0131n da\u011f\u0131t\u0131m s\u0131ras\u0131nda bile kesintisiz \u00e7al\u0131\u015fmas\u0131, kullan\u0131c\u0131lar\u0131n g\u00fcvenini art\u0131r\u0131r ve onlara s\u00fcrekli eri\u015filebilir bir hizmet sundu\u011funuzu g\u00f6sterir. Kesintiler, kullan\u0131c\u0131lar\u0131n hayal k\u0131r\u0131kl\u0131\u011f\u0131na u\u011framas\u0131na, sitenizden ayr\u0131lmas\u0131na ve hatta rakip platformlara y\u00f6nelmesine neden olabilir. S\u0131f\u0131r kesinti da\u011f\u0131t\u0131m\u0131, bu riski ortadan kald\u0131rarak kullan\u0131c\u0131 sadakatini peki\u015ftirir.<\/p>\n
E-ticaret siteleri, finansal uygulamalar veya kritik i\u015f s\u00fcre\u00e7lerini y\u00f6neten platformlar i\u00e7in her saniye \u00f6nemlidir. Bir kesinti, do\u011frudan gelir kayb\u0131na, m\u00fc\u015fteri kayb\u0131na ve hatta yasal y\u00fck\u00fcml\u00fcl\u00fcklere yol a\u00e7abilir. S\u0131f\u0131r kesinti da\u011f\u0131t\u0131m\u0131, i\u015f s\u00fcre\u00e7lerinizin aksamadan devam etmesini sa\u011flayarak potansiyel zararlar\u0131 minimize eder.<\/p>\n
Pazar liderleri, genellikle en yeni \u00f6zellikleri en h\u0131zl\u0131 ve en sorunsuz \u015fekilde sunanlard\u0131r. S\u0131f\u0131r kesinti yetene\u011fi, ekiplerin daha s\u0131k da\u011f\u0131t\u0131m yapmas\u0131na olanak tan\u0131r, bu da \u00fcr\u00fcn\u00fcn daha h\u0131zl\u0131 iterasyonunu ve pazar de\u011fi\u015fikliklerine daha \u00e7evik yan\u0131t vermesini sa\u011flar. Bu \u00e7eviklik, rakiplerinize kar\u015f\u0131 \u00f6nemli bir avantaj sa\u011flar.<\/p>\n
Arama motorlar\u0131, bir sitenin kullan\u0131labilirli\u011fini s\u0131ralama fakt\u00f6rlerinden biri olarak de\u011ferlendirir. S\u0131k ya\u015fanan kesintiler veya uzun s\u00fcreli eri\u015filemezlik, arama motoru s\u0131ralaman\u0131z\u0131 olumsuz etkileyebilir. S\u0131f\u0131r kesinti da\u011f\u0131t\u0131m\u0131, sitenizin s\u00fcrekli olarak taranabilir ve eri\u015filebilir olmas\u0131n\u0131 sa\u011flayarak SEO performans\u0131n\u0131z\u0131 koruman\u0131za yard\u0131mc\u0131 olur.<\/p>\n
S\u0131f\u0131r kesinti Rails da\u011f\u0131t\u0131m\u0131n\u0131 m\u00fcmk\u00fcn k\u0131lmak i\u00e7in birka\u00e7 temel bile\u015fenin uyumlu bir \u015fekilde \u00e7al\u0131\u015fmas\u0131 gerekir. Bu bile\u015fenler, uygulaman\u0131n farkl\u0131 katmanlar\u0131nda g\u00f6rev alarak kesintisiz ge\u00e7i\u015fi sa\u011flar.<\/p>\n
Puma, Ruby on Rails uygulamalar\u0131 i\u00e7in y\u00fcksek performansl\u0131, \u00e7ok i\u015f par\u00e7ac\u0131kl\u0131 ve \u00e7ok i\u015flemli bir HTTP sunucusudur. \u00d6zellikle “cluster mode” ad\u0131 verilen \u00e7ok i\u015flemli \u00e7al\u0131\u015fma yetene\u011fi, s\u0131f\u0131r kesinti da\u011f\u0131t\u0131mlar\u0131 i\u00e7in kilit bir rol oynar. Puma, yeni kod da\u011f\u0131t\u0131ld\u0131\u011f\u0131nda eski uygulama \u00f6rneklerini zarif bir \u015fekilde kapat\u0131rken yeni \u00f6rnekleri ba\u015flatma yetene\u011fine sahiptir. Bu “hot restart” veya “phased restart” \u00f6zelli\u011fi, mevcut istekleri tamamlamas\u0131na izin verirken yeni gelen istekleri yeni kod taban\u0131na y\u00f6nlendirir.<\/p>\n
Foreman, Procfile tabanl\u0131 uygulamalar\u0131 y\u00f6netmek i\u00e7in kullan\u0131lan bir ara\u00e7t\u0131r. Bir Bir ters proxy (reverse proxy), istemciden gelen istekleri al\u0131r ve bunlar\u0131 arka u\u00e7taki uygulama sunucular\u0131na (bizim durumumuzda Puma) iletir. Nginx veya Apache gibi pop\u00fcler ters proxy’ler, y\u00fck dengeleme, SSL sonland\u0131rma, statik dosya sunumu ve en \u00f6nemlisi, da\u011f\u0131t\u0131m s\u0131ras\u0131nda arka u\u00e7 uygulama sunucular\u0131n\u0131 de\u011fi\u015ftirmek i\u00e7in kullan\u0131l\u0131r. Nginx, yeni bir uygulama s\u00fcr\u00fcm\u00fc da\u011f\u0131t\u0131ld\u0131\u011f\u0131nda, eski Puma \u00f6rnekleri kapan\u0131rken yeni ba\u015flat\u0131lan Puma \u00f6rneklerine sorunsuz bir \u015fekilde trafik y\u00f6nlendirme yetene\u011fi sayesinde s\u0131f\u0131r kesinti da\u011f\u0131t\u0131mlar\u0131n\u0131n kritik bir par\u00e7as\u0131d\u0131r.<\/p>\n Capistrano, sunuculara otomatik olarak kod da\u011f\u0131tmak i\u00e7in kullan\u0131lan pop\u00fcler bir uzaktan sunucu otomasyon arac\u0131d\u0131r. Karma\u015f\u0131k da\u011f\u0131t\u0131m ad\u0131mlar\u0131n\u0131 (kod \u00e7ekme, ba\u011f\u0131ml\u0131l\u0131klar\u0131 y\u00fckleme, veritaban\u0131 migrasyonlar\u0131, sunucu yeniden ba\u015flatma) otomatikle\u015ftirmeyi sa\u011flar. S\u0131f\u0131r kesinti da\u011f\u0131t\u0131m\u0131, manuel ad\u0131mlarla y\u00f6netilemeyecek kadar karma\u015f\u0131k oldu\u011fundan, Capistrano gibi bir arac\u0131n kullan\u0131lmas\u0131 zorunludur. Capistrano, \u00f6zellikle Puma’n\u0131n yeniden ba\u015flatma sinyallerini do\u011fru zamanda g\u00f6ndermek ve Nginx yap\u0131land\u0131rmas\u0131n\u0131 g\u00fcncelleyerek trafi\u011fi yeni uygulama \u00f6rne\u011fine y\u00f6nlendirmek i\u00e7in \u00f6zelle\u015ftirilebilir.<\/p>\n S\u0131f\u0131r kesinti da\u011f\u0131t\u0131m mimarisi, uygulaman\u0131z\u0131n kesintisiz \u00e7al\u0131\u015fmas\u0131n\u0131 sa\u011flamak i\u00e7in katmanl\u0131 bir yakla\u015f\u0131ma dayan\u0131r. Bu mimarinin kalbinde, istemci isteklerini uygulama sunucular\u0131na y\u00f6nlendiren bir ters proxy ve uygulaman\u0131n kendisini bar\u0131nd\u0131ran \u00e7ok i\u015flemli bir uygulama sunucusu bulunur.<\/p>\n Genel olarak, bir istemci (kullan\u0131c\u0131 web taray\u0131c\u0131s\u0131), uygulaman\u0131za bir istek g\u00f6nderdi\u011finde, bu istek \u00f6nce ters proxy sunucusuna (\u00f6rne\u011fin Nginx) ula\u015f\u0131r. Nginx, bu iste\u011fi al\u0131r ve \u00f6nceden yap\u0131land\u0131r\u0131lm\u0131\u015f kurallara g\u00f6re arka u\u00e7taki uygulama sunucular\u0131ndan birine (Puma) iletir. Puma, Rails uygulamas\u0131n\u0131 \u00e7al\u0131\u015ft\u0131r\u0131r, iste\u011fi i\u015fler ve yan\u0131t\u0131 Nginx arac\u0131l\u0131\u011f\u0131yla istemciye geri g\u00f6nderir.<\/p>\n Da\u011f\u0131t\u0131m s\u0131ras\u0131nda, bu ak\u0131\u015f kritik bir d\u00f6n\u00fc\u015f\u00fcme u\u011frar. Yeni kod sunucuya \u00e7ekilir, ba\u011f\u0131ml\u0131l\u0131klar y\u00fcklenir ve yeni uygulama s\u00fcr\u00fcm\u00fc haz\u0131rlan\u0131r. Ancak bu yeni s\u00fcr\u00fcm hemen eski s\u00fcr\u00fcm\u00fcn yerini almaz. Bunun yerine, Puma’n\u0131n “hot restart” yetene\u011fi devreye girer. Eski Puma i\u015flemleri hala \u00e7al\u0131\u015f\u0131r durumdayken, yeni kodla birlikte yeni Puma i\u015flemleri ba\u015flat\u0131l\u0131r. Nginx, eski Puma’n\u0131n soketine ba\u011fl\u0131 kalmaya devam ederken, yeni Puma i\u015flemleri kendi yeni soketlerinde dinlemeye ba\u015flar. Puma’n\u0131n yeniden ba\u015flatma sinyali (SIGUSR2) al\u0131nd\u0131\u011f\u0131nda, Nginx’e yeni soket bilgisi bildirilir ve Nginx yava\u015f\u00e7a yeni Puma soketine trafik y\u00f6nlendirmeye ba\u015flar. Eski Puma i\u015flemleri, mevcut isteklerini tamamlad\u0131ktan sonra zarif bir \u015fekilde kapan\u0131r. Bu ge\u00e7i\u015f, kullan\u0131c\u0131lar i\u00e7in tamamen \u015feffaf bir \u015fekilde ger\u00e7ekle\u015fir, b\u00f6ylece herhangi bir kesinti ya\u015fanmaz.<\/p>\n Puma’n\u0131n s\u0131f\u0131r kesinti da\u011f\u0131t\u0131mlar\u0131ndaki rol\u00fc, onun \u00e7oklu i\u015flem yeteneklerine ve sinyal i\u015fleme mekanizmas\u0131na dayan\u0131r. Bu \u00f6zellikler, yeni bir kod s\u00fcr\u00fcm\u00fcn\u00fc etkinle\u015ftirirken eski s\u00fcr\u00fcm\u00fcn hala aktif istekleri i\u015flemesine olanak tan\u0131r.<\/p>\n Puma, birden fazla worker i\u015flemi \u00e7al\u0131\u015ft\u0131rarak Rails uygulaman\u0131z\u0131n performans\u0131n\u0131 ve kararl\u0131l\u0131\u011f\u0131n\u0131 art\u0131r\u0131r. Her worker i\u015flemi, kendi Rails uygulama \u00f6rne\u011fini bar\u0131nd\u0131r\u0131r ve birden fazla i\u015f par\u00e7ac\u0131\u011f\u0131 (thread) ile e\u015fzamanl\u0131 istekleri i\u015fleyebilir. Cluster mode, bu worker’lar\u0131n ana bir Puma master i\u015flemi taraf\u0131ndan y\u00f6netilmesini sa\u011flar. Master i\u015flem, worker’lar\u0131 ba\u015flatmaktan, izlemekten ve gerekti\u011finde yeniden ba\u015flatmaktan sorumludur. S\u0131f\u0131r kesinti da\u011f\u0131t\u0131m\u0131nda, bu master i\u015flem, yeni kod da\u011f\u0131t\u0131ld\u0131\u011f\u0131nda worker’lar\u0131 g\u00fcvenli bir \u015fekilde g\u00fcncellemeyi koordine eder.<\/p>\n Puma’n\u0131n “hot restart” veya “phased restart” mekanizmas\u0131, s\u0131f\u0131r kesintinin anahtar\u0131d\u0131r. Bu mekanizma, a\u015fa\u011f\u0131daki sinyaller arac\u0131l\u0131\u011f\u0131yla tetiklenir:<\/p>\n * Bu s\u00fcre\u00e7 sayesinde, uygulaman\u0131z\u0131n her zaman en az bir worker k\u00fcmesi taraf\u0131ndan hizmet verilmesi sa\u011flan\u0131r. Kullan\u0131c\u0131lar, da\u011f\u0131t\u0131m\u0131n ger\u00e7ekle\u015fti\u011fini fark etmezler \u00e7\u00fcnk\u00fc istekleri her zaman aktif ve do\u011fru bir uygulama \u00f6rne\u011fi taraf\u0131ndan kar\u015f\u0131lan\u0131r.<\/p>\n Foreman, geli\u015ftirme ortam\u0131nda Procfile tabanl\u0131 uygulamalar\u0131 y\u00f6netmek i\u00e7in tasarlanm\u0131\u015f olsa da, \u00fcretim ortam\u0131nda da benzer bir i\u015flevsellik sa\u011flayan ara\u00e7larla birlikte kullan\u0131labilir.<\/p>\n Bir \u00d6rnek bir Bu \u00f6rnekte: Foreman, Do\u011frudan Foreman’\u0131 \u00fcretimde kullanmak yerine, genellikle * S\u0131f\u0131r kesinti da\u011f\u0131t\u0131m\u0131nda, bu \u00fcretim s\u00fcre\u00e7 y\u00f6neticileri, Capistrano gibi da\u011f\u0131t\u0131m ara\u00e7lar\u0131ndan gelen sinyallere yan\u0131t vererek Puma’n\u0131n hot restart mekanizmas\u0131n\u0131 do\u011fru bir \u015fekilde tetiklemekten sorumludur.<\/p>\n Capistrano ile s\u0131f\u0131r kesinti da\u011f\u0131t\u0131m\u0131, bir dizi dikkatlice orkestrasyon edilmi\u015f ad\u0131mdan olu\u015fur. Bu ad\u0131mlar, yeni kodun g\u00fcvenli bir \u015fekilde sunucuya aktar\u0131lmas\u0131n\u0131 ve uygulaman\u0131n kesintisiz bir \u015fekilde g\u00fcncellenmesini sa\u011flar.<\/p>\n Da\u011f\u0131t\u0131ma ba\u015flamadan \u00f6nce sunucunuzun haz\u0131r oldu\u011fundan emin olmal\u0131s\u0131n\u0131z:<\/p>\n * Sunucu \u0130\u015fletim Sistemi<\/strong>: Ubuntu, CentOS gibi Linux da\u011f\u0131t\u0131mlar\u0131. 1. Ard\u0131ndan 2. 3. 4. 5. Capistrano’nun 1. 2. Nginx, istemci isteklerini Puma’ya y\u00f6nlendiren ters proxy g\u00f6revi g\u00f6r\u00fcr. Bu yap\u0131land\u0131rma, Nginx’in T\u00fcm yap\u0131land\u0131rmalar tamamland\u0131ktan sonra, da\u011f\u0131t\u0131m\u0131 ba\u015flatmak i\u00e7in a\u015fa\u011f\u0131daki komutu kullan\u0131rs\u0131n\u0131z:<\/p>\n Bu komutun arkas\u0131nda Capistrano, bir dizi g\u00f6revi belirli bir s\u0131rayla \u00e7al\u0131\u015ft\u0131r\u0131r:<\/p>\n 1. Bu s\u00fcre\u00e7 boyunca, Nginx her zaman \u00e7al\u0131\u015fan bir Puma \u00f6rne\u011fine trafik y\u00f6nlendirdi\u011fi i\u00e7in kullan\u0131c\u0131lar herhangi bir kesinti ya\u015famazlar.<\/p>\n Veritaban\u0131 migrasyonlar\u0131, s\u0131f\u0131r kesinti da\u011f\u0131t\u0131mlar\u0131n\u0131n en zorlu k\u0131s\u0131mlar\u0131ndan biridir. Yeni kod da\u011f\u0131t\u0131l\u0131rken, eski kod hala aktif oldu\u011fundan, migrasyonlar\u0131n her iki kod s\u00fcr\u00fcm\u00fcyle de uyumlu olmas\u0131 gerekir.<\/p>\n * Non-blocking Migrasyonlar<\/strong>: Uzun s\u00fcren S\u0131f\u0131r kesinti da\u011f\u0131t\u0131mlar\u0131, bir\u00e7ok avantaj sunsa da, dikkatli planlama ve uygulama gerektiren potansiyel sorunlara sahiptir.<\/p>\n * Uzun S\u00fcren Veritaban\u0131 Migrasyonlar\u0131<\/strong>: Yukar\u0131da belirtildi\u011fi gibi, uzun s\u00fcren migrasyonlar tablo kilitlenmelerine ve kesintiye yol a\u00e7abilir. \u00c7\u00f6z\u00fcm, Puma ve Capistrano ile s\u0131f\u0131r kesinti da\u011f\u0131t\u0131m\u0131 yayg\u0131n bir y\u00f6ntem olsa da, ba\u015fka g\u00fc\u00e7l\u00fc alternatifler de mevcuttur:<\/p>\n * Docker ve Kubernetes<\/strong>: Konteyner teknolojileri, uygulamalar\u0131 izole edilmi\u015f ortamlarda \u00e7al\u0131\u015ft\u0131rmay\u0131 ve \u00f6l\u00e7eklendirmeyi kolayla\u015ft\u0131r\u0131r. Kubernetes, da\u011f\u0131t\u0131mlar\u0131, \u00f6l\u00e7eklendirmeyi ve hizmet ke\u015ffini otomatikle\u015ftiren bir konteyner orkestrasyon platformudur. Blue\/Green veya Canary da\u011f\u0131t\u0131mlar\u0131 gibi geli\u015fmi\u015f s\u0131f\u0131r kesinti stratejilerini Kubernetes \u00fczerinde uygulamak daha kolayd\u0131r. S\u0131f\u0131r kesinti Rails da\u011f\u0131t\u0131mlar\u0131, modern web uygulamalar\u0131 i\u00e7in vazge\u00e7ilmez bir uygulamad\u0131r. Kullan\u0131c\u0131 deneyimini iyile\u015ftirir, i\u015f s\u00fcreklili\u011fini sa\u011flar ve ekiplerin daha \u00e7evik olmas\u0131na olanak tan\u0131r. Puma’n\u0131n hot restart yetenekleri, Foreman’\u0131n s\u00fcre\u00e7 y\u00f6netimi ve Capistrano’nun otomasyon g\u00fcc\u00fc bir araya geldi\u011finde, Rails uygulamalar\u0131n\u0131z\u0131 kesintisiz bir \u015fekilde g\u00fcncelleyebilece\u011finiz sa\u011flam bir mekanizma olu\u015fturur.<\/p>\n Bu s\u00fcre\u00e7, do\u011fru yap\u0131land\u0131rma ve veritaban\u0131 migrasyonlar\u0131na \u00f6zel dikkat gerektirse de, uzun vadede uygulaman\u0131z\u0131n kararl\u0131l\u0131\u011f\u0131 ve kullan\u0131c\u0131 memnuniyeti i\u00e7in yap\u0131lan yat\u0131r\u0131m\u0131n kar\u015f\u0131l\u0131\u011f\u0131n\u0131 fazlas\u0131yla verir. Alternatif teknolojiler ve stratejiler de mevcut olsa da, Puma ve Capistrano kombinasyonu, bir\u00e7ok Rails geli\u015ftiricisi i\u00e7in eri\u015filebilir ve g\u00fc\u00e7l\u00fc bir s\u0131f\u0131r kesinti da\u011f\u0131t\u0131m \u00e7\u00f6z\u00fcm\u00fc sunmaya devam etmektedir. Uygulaman\u0131z\u0131n ve ekibinizin ihtiya\u00e7lar\u0131na en uygun yakla\u015f\u0131m\u0131 se\u00e7mek, ba\u015far\u0131l\u0131 bir s\u0131f\u0131r kesinti da\u011f\u0131t\u0131m stratejisinin anahtar\u0131d\u0131r.<\/body><\/p>\n","protected":false},"excerpt":{"rendered":"Puma ve Foreman Kullanarak S\u0131f\u0131r Kesintiyle Rails Da\u011f\u0131t\u0131mlar\u0131 Nas\u0131l Yap\u0131l\u0131r?\nS\u00fcrekli geli\u015fen web uygulamalar\u0131 d\u00fcnyas\u0131nda, kullan\u0131c\u0131 d","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"csco_page_header_type":"","csco_page_load_nextpost":"","csco_page_subscribe_form":"","csco_page_contact_form":"","footnotes":""},"categories":[1342],"tags":[],"class_list":{"0":"post-34204","1":"post","2":"type-post","3":"status-publish","4":"format-standard","6":"category-ai","7":"cs-entry","8":"cs-video-wrap"},"yoast_head":"\nProcfile<\/code>, uygulaman\u0131z\u0131n farkl\u0131 s\u00fcre\u00e7lerini (web sunucusu, worker’lar, zamanlanm\u0131\u015f g\u00f6revler vb.) tan\u0131mlayan basit bir metin dosyas\u0131d\u0131r. Foreman, bu s\u00fcre\u00e7leri yerel geli\u015ftirme ortam\u0131nda veya \u00fcretimde ba\u015flatman\u0131za, durdurman\u0131za ve izlemenize olanak tan\u0131r. S\u0131f\u0131r kesinti da\u011f\u0131t\u0131m senaryolar\u0131nda, Foreman veya onun \u00fcretim ortam\u0131ndaki kar\u015f\u0131l\u0131klar\u0131 (systemd, monit, god) Puma s\u00fcre\u00e7lerini do\u011fru sinyallerle y\u00f6netmek i\u00e7in kullan\u0131l\u0131r.<\/p>\nReverse Proxy (Nginx\/Apache)<\/h4>\n
Capistrano (veya Benzeri Bir Da\u011f\u0131t\u0131m Arac\u0131)<\/h4>\n
S\u0131f\u0131r Kesinti Da\u011f\u0131t\u0131m Mimarisi<\/h3>\n
Puma ile S\u0131f\u0131r Kesinti Nas\u0131l Sa\u011flan\u0131r?<\/h3>\n
Puma’n\u0131n \u00c7oklu \u0130\u015flem Modu (Cluster Mode)<\/h4>\n
Puma’n\u0131n Hot Restart Mekanizmas\u0131<\/h4>\n
SIGTERM<\/code><\/strong>: Bu sinyal, Puma’ya mevcut istekleri tamamlad\u0131ktan sonra zarif bir \u015fekilde kapanmas\u0131n\u0131 s\u00f6yler. Genellikle uygulaman\u0131n tamamen durdurulmas\u0131 gerekti\u011finde kullan\u0131l\u0131r.
\n* SIGUSR2<\/code><\/strong>: Bu sinyal, Puma master i\u015flemine “hot restart” yapmas\u0131n\u0131 emreder. Master i\u015flem bu sinyali ald\u0131\u011f\u0131nda:
\n 1. Yeni kod taban\u0131yla yeni worker i\u015flemleri ba\u015flat\u0131r.
\n 2. Yeni worker’lar ba\u015flat\u0131l\u0131p haz\u0131r hale geldi\u011finde, master i\u015flem eski worker’lara SIGTERM<\/code> sinyali g\u00f6nderir.
\n 3. Eski worker’lar, mevcut isteklerini tamamlad\u0131ktan sonra kapan\u0131r.
\n 4. Nginx gibi ters proxy’ler, bu s\u00fcre\u00e7te yeni Puma soketine y\u00f6nlendirilerek trafi\u011fi kesintisiz bir \u015fekilde yeni worker’lara aktar\u0131r.<\/p>\nForeman ile S\u00fcre\u00e7 Y\u00f6netimi<\/h3>\n
Procfile<\/code> Nedir ve Nas\u0131l Olu\u015fturulur?<\/h4>\nProcfile<\/code>, uygulaman\u0131z\u0131n hangi s\u00fcre\u00e7lerden olu\u015ftu\u011funu ve bu s\u00fcre\u00e7lerin nas\u0131l ba\u015flat\u0131lmas\u0131 gerekti\u011fini tan\u0131mlayan basit bir metin dosyas\u0131d\u0131r. Genellikle uygulaman\u0131z\u0131n k\u00f6k dizininde bulunur.<\/p>\nProcfile<\/code>:<\/p>\nweb: bundle exec puma -C config\/puma.rb\nworker: bundle exec sidekiq -c 5 -q default -q mailers<\/code><\/pre>\n
\n* web<\/code>: Puma web sunucusunu config\/puma.rb<\/code> yap\u0131land\u0131rma dosyas\u0131yla ba\u015flat\u0131r.
\n* worker<\/code>: Sidekiq arka plan i\u015fleme worker’\u0131n\u0131 ba\u015flat\u0131r.<\/p>\nForeman’\u0131n G\u00f6revi<\/h4>\n
Procfile<\/code>‘\u0131 okur ve tan\u0131mlanan her s\u00fcre\u00e7 i\u00e7in ayr\u0131 bir i\u015flem ba\u015flat\u0131r. Bu, \u00f6zellikle yerel geli\u015ftirme ortam\u0131nda uygulaman\u0131z\u0131n t\u00fcm bile\u015fenlerini tek bir komutla (\u00f6rne\u011fin foreman start<\/code>) ba\u015flatman\u0131za olanak tan\u0131r. Her s\u00fcrecin \u00e7\u0131kt\u0131s\u0131n\u0131 ayr\u0131 pencerelerde veya tek bir konsolda birle\u015ftirerek izlemeyi kolayla\u015ft\u0131r\u0131r.<\/p>\n\u00dcretim Ortam\u0131nda Foreman Kullan\u0131m\u0131<\/h4>\n
Procfile<\/code>‘\u0131 okuyup sisteminize \u00f6zel s\u00fcre\u00e7 y\u00f6neticisi yap\u0131land\u0131rmalar\u0131 \u00fcreten ara\u00e7lar tercih edilir. \u00d6rne\u011fin:<\/p>\nsystemd<\/code><\/strong>: Linux sistemlerinde birincil init sistemi ve servis y\u00f6neticisidir. systemd<\/code> servis dosyalar\u0131, Procfile<\/code>‘daki her bir s\u00fcre\u00e7 i\u00e7in olu\u015fturularak Puma ve di\u011fer uygulama bile\u015fenlerinin sistem ba\u015flang\u0131c\u0131nda otomatik olarak ba\u015flat\u0131lmas\u0131n\u0131 ve izlenmesini sa\u011flar. systemd-foreman<\/code> veya foreman export systemd<\/code> gibi ara\u00e7lar bu d\u00f6n\u00fc\u015f\u00fcm\u00fc yapabilir.
\n* monit<\/code> \/ god<\/code><\/strong>: Bu ara\u00e7lar, s\u00fcre\u00e7leri izlemek, \u00e7\u00f6kt\u00fcklerinde otomatik olarak yeniden ba\u015flatmak ve belirli ko\u015fullar alt\u0131nda bildirimler g\u00f6ndermek i\u00e7in kullan\u0131l\u0131r. Puma gibi kritik s\u00fcre\u00e7lerin her zaman \u00e7al\u0131\u015f\u0131r durumda kalmas\u0131n\u0131 sa\u011flamak i\u00e7in faydal\u0131d\u0131rlar.<\/p>\nDa\u011f\u0131t\u0131m S\u00fcreci Ad\u0131m Ad\u0131m (Capistrano ile \u00d6rnek)<\/h3>\n
\u00d6n Gereksinimler<\/h4>\n
\n* SSH Anahtarlar\u0131<\/strong>: Capistrano’nun sunucuya g\u00fcvenli bir \u015fekilde ba\u011flanabilmesi i\u00e7in SSH anahtar tabanl\u0131 kimlik do\u011frulama.
\n* Ruby ve Rails Ortam\u0131<\/strong>: rbenv<\/code> veya rvm<\/code> ile do\u011fru Ruby s\u00fcr\u00fcm\u00fc ve Bundler.
\n* Veritaban\u0131<\/strong>: PostgreSQL veya MySQL kurulu ve yap\u0131land\u0131r\u0131lm\u0131\u015f olmal\u0131.
\n* Ters Proxy<\/strong>: Nginx kurulu ve yap\u0131land\u0131r\u0131lm\u0131\u015f olmal\u0131.
\n* Uygulama Ba\u011f\u0131ml\u0131l\u0131klar\u0131<\/strong>: git<\/code>, nodejs<\/code>, yarn<\/code> (asset derleme i\u00e7in).<\/p>\nCapistrano Kurulumu ve Yap\u0131land\u0131rmas\u0131<\/h4>\n
Gemfile<\/code>‘a Ekleme<\/strong>:<\/p>\ngroup :development do\n gem 'capistrano', '~> 3.17'\n gem 'capistrano-rails'\n gem 'capistrano-rbenv', '~> 2.1' # rbenv kullan\u0131yorsan\u0131z\n gem 'capistrano-puma', '~> 6.0' # Puma i\u00e7in\n gem 'capistrano-nginx', '~> 2.0' # Nginx i\u00e7in (iste\u011fe ba\u011fl\u0131, manuel konfig\u00fcrasyon da m\u00fcmk\u00fcn)\n end<\/code><\/pre>\nbundle install<\/code> \u00e7al\u0131\u015ft\u0131r\u0131n.<\/p>\ncap install<\/code><\/strong>: Bu komut, Capistrano’nun temel yap\u0131land\u0131rma dosyalar\u0131n\u0131 (\u00f6rne\u011fin Capfile<\/code>, config\/deploy.rb<\/code>, config\/deploy\/production.rb<\/code>) olu\u015fturur.<\/p>\nconfig\/deploy.rb<\/code> Yap\u0131land\u0131rmas\u0131<\/strong>: Bu dosya, t\u00fcm da\u011f\u0131t\u0131m ortamlar\u0131 i\u00e7in ortak ayarlar\u0131 i\u00e7erir.<\/p>\n# config\/deploy.rb\n lock \"~> 3.17.3\"\n\n set :application, \"your_app_name\"\n set :repo_url, \"git@github.com:your_user\/your_app_name.git\"\n\n # Default branch is :master\n # ask :branch, git rev-parse --abbrev-ref HEAD<\/code>.chomp\n set :branch, ENV['BRANCH'] || 'master'\n\n # Default deploy_to directory is \/var\/www\/my_app_name\n set :deploy_to, \"\/var\/www\/#{fetch(:application)}\"\n\n # Default value for :format is :airbrussh.\n # set :format, :airbrussh\n\n # You can configure the Airbrussh format using :format_options.\n # These are the defaults.\n # set :format_options, command_output: true, log_file: \"log\/capistrano.log\", color: :auto, truncate: :auto\n\n # Default value for :pty is false\n set :pty, true\n\n # Default value for :linked_files is []\n append :linked_files, \"config\/database.yml\", \"config\/master.key\" # master.key Rails 5.2+ i\u00e7in\n # append :linked_files, \"config\/credentials\/#{fetch(:stage)}.yml.enc\" # Rails 6+ i\u00e7in\n\n # Default value for linked_dirs is []\n append :linked_dirs, \"log\", \"tmp\/pids\", \"tmp\/cache\", \"tmp\/sockets\", \"public\/system\", \"vendor\/bundle\", \"public\/uploads\"\n\n # Default value for default_env is {}\n # set :default_env, { path: \"\/opt\/ruby\/bin:$PATH\" }\n\n # Default value for local_user is ENV['USER']\n # set :local_user, -> { git config user.name<\/code>.chomp }\n\n # Default value for keep_releases is 5\n set :keep_releases, 5\n\n # Uncomment the following to require manually confirming clear\/revert actions when running deploy:clear_cache<\/code> or deploy:revert<\/code>.\n # set :confirm_clear, true\n\n # rbenv configuration\n set :rbenv_type, :user # or :system, depends on your rbenv installation\n set :rbenv_ruby, '3.2.2' # Your Ruby version\n set :rbenv_prefix, \"RBENV_ROOT=#{fetch(:rbenv_path)} RBENV_VERSION=#{fetch(:rbenv_ruby)} #{fetch(:rbenv_path)}\/bin\/rbenv exec\"\n set :rbenv_map_bins, %w{rake gem bundle ruby rails puma pumactl sidekiq sidekiqctl}\n set :rbenv_roles, :all # default value\n\n # Puma configuration\n set :puma_init_active_record, true # For phased restart, ensures DB connection is handled gracefully<\/code><\/pre>\nlinked_files<\/code> ve linked_dirs<\/code>, her da\u011f\u0131t\u0131mda sembolik olarak ba\u011flanan dosyalar\u0131 ve dizinleri belirtir. Bu, da\u011f\u0131t\u0131mlar aras\u0131nda kal\u0131c\u0131 olmas\u0131 gereken veriler (veritaban\u0131 yap\u0131land\u0131rmas\u0131, loglar, y\u00fcklemeler) i\u00e7in \u00f6nemlidir.<\/p>\nconfig\/deploy\/production.rb<\/code> Yap\u0131land\u0131rmas\u0131<\/strong>: Ortama \u00f6zel ayarlar buraya yaz\u0131l\u0131r.<\/p>\n# config\/deploy\/production.rb\n server 'your_server_ip_or_hostname', user: 'deploy_user', roles: %w{web app db}\n\n set :rails_env, :production\n set :stage, :production<\/code><\/pre>\nshared<\/code> Dizini<\/strong>: Capistrano, deploy_to<\/code> dizini alt\u0131nda releases<\/code> ve shared<\/code> ad\u0131nda iki ana dizin olu\u015fturur. releases<\/code> dizini, uygulaman\u0131z\u0131n her yeni s\u00fcr\u00fcm\u00fcn\u00fc i\u00e7erirken, shared<\/code> dizini, t\u00fcm s\u00fcr\u00fcmler aras\u0131nda payla\u015f\u0131lmas\u0131 gereken kal\u0131c\u0131 dosyalar\u0131 ve dizinleri (\u00f6rne\u011fin config\/database.yml<\/code>, log<\/code>, tmp\/pids<\/code>) bar\u0131nd\u0131r\u0131r. Da\u011f\u0131t\u0131m s\u0131ras\u0131nda, Capistrano shared<\/code> dizinindeki bu dosya ve dizinlere sembolik ba\u011flant\u0131lar olu\u015fturur.<\/p>\nPuma Yap\u0131land\u0131rmas\u0131 (Capistrano ile)<\/h4>\n
capistrano-puma<\/code> gem’i, Puma’n\u0131n da\u011f\u0131t\u0131m\u0131n\u0131 ve y\u00f6netimini kolayla\u015ft\u0131r\u0131r.<\/p>\nconfig\/puma.rb<\/code> (Uygulama \u0130\u00e7in)<\/strong>: Bu dosya, Puma’n\u0131n nas\u0131l \u00e7al\u0131\u015faca\u011f\u0131n\u0131 tan\u0131mlar.<\/p>\n# config\/puma.rb\n workers Integer(ENV['WEB_CONCURRENCY'] || 2) # Worker say\u0131s\u0131, genellikle CPU \u00e7ekirdek say\u0131s\u0131na g\u00f6re ayarlan\u0131r\n threads Integer(ENV['MIN_THREADS'] || 1), Integer(ENV['MAX_THREADS'] || 16) # \u0130\u015f par\u00e7ac\u0131\u011f\u0131 say\u0131s\u0131\n\n preload_app! # Uygulamay\u0131 worker'lar fork edilmeden \u00f6nce y\u00fckle (daha h\u0131zl\u0131 restart)\n\n rackup DefaultRackup\n port ENV['PORT'] || 3000\n environment ENV['RACK_ENV'] || 'development'\n\n on_worker_boot do\n # Worker ba\u015flat\u0131ld\u0131\u011f\u0131nda ActiveRecord ba\u011flant\u0131lar\u0131n\u0131 yeniden kur\n ActiveRecord::Base.establish_connection if defined?(ActiveRecord)\n end\n\n # production ortam\u0131 i\u00e7in socket ba\u011flant\u0131s\u0131\n # Capistrano-puma otomatik olarak bu dosyay\u0131 olu\u015fturur ve y\u00f6netir.\n # set :puma_bind, \"unix:\/\/#{shared_path}\/tmp\/sockets\/puma.sock\"\n # set :puma_state, \"#{shared_path}\/tmp\/pids\/puma.state\"\n # set :puma_pid, \"#{shared_path}\/tmp\/pids\/puma.pid\"<\/code><\/pre>\npreload_app!<\/code> direktifi, master Puma i\u015fleminin uygulamay\u0131 y\u00fcklemesini sa\u011flar, b\u00f6ylece worker’lar fork edildi\u011finde uygulama kodu zaten bellekte olur. Bu, hot restart s\u0131ras\u0131nda daha h\u0131zl\u0131 ba\u015flang\u0131\u00e7 sa\u011flar.<\/p>\ncapistrano-puma<\/code> Ayarlar\u0131<\/strong>: deploy.rb<\/code> i\u00e7inde Capistrano-Puma i\u00e7in ek ayarlar yapabilirsiniz.<\/p>\n# config\/deploy.rb\n # ...\n set :puma_user, fetch(:user) # Puma'n\u0131n \u00e7al\u0131\u015faca\u011f\u0131 kullan\u0131c\u0131\n set :puma_rackup, -> { File.join(current_path, 'config.ru') }\n set :puma_state, \"#{shared_path}\/tmp\/pids\/puma.state\"\n set :puma_pid, \"#{shared_path}\/tmp\/pids\/puma.pid\"\n set :puma_bind, \"unix:\/\/#{shared_path}\/tmp\/sockets\/puma.sock\" # Nginx'in ba\u011flanaca\u011f\u0131 soket\n set :puma_default_control_app, \"unix:\/\/#{shared_path}\/tmp\/sockets\/pumactl.sock\"\n set :puma_access_log, \"#{shared_path}\/log\/puma_access.log\"\n set :puma_error_log, \"#{shared_path}\/log\/puma_error.log\"\n set :puma_worker_timeout, nil # Puma worker'lar\u0131n\u0131n zaman a\u015f\u0131m\u0131\n set :puma_preload_app, true # preload_app! kullan\u0131yorsan\u0131z true yap\u0131n\n set :puma_init_active_record, true # hot restart s\u0131ras\u0131nda ActiveRecord ba\u011flant\u0131lar\u0131n\u0131 yeniler\n # ...<\/code><\/pre>\nNginx Yap\u0131land\u0131rmas\u0131<\/h4>\n
capistrano-nginx<\/code> gem’i, Nginx yap\u0131land\u0131rmas\u0131n\u0131 otomatik olarak olu\u015fturabilir, ancak manuel olarak da yapabilirsiniz.<\/p>\n# \/etc\/nginx\/sites-available\/your_app_name.conf\nupstream puma {\n server unix:\/var\/www\/your_app_name\/shared\/tmp\/sockets\/puma.sock fail_timeout=0;\n}\n\nserver {\n listen 80;\n listen [::]:80;\n server_name your_domain.com www.your_domain.com;\n\n root \/var\/www\/your_app_name\/current\/public;\n\n try_files $uri\/index.html $uri @puma;\n\n location @puma {\n proxy_pass http:\/\/puma;\n proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;\n proxy_set_header Host $http_host;\n proxy_redirect off;\n }\n\n error_page 500 502 503 504 \/500.html;\n client_max_body_size 10M;\n keepalive_timeout 10;\n}<\/code><\/pre>\n\/var\/www\/your_app_name\/shared\/tmp\/sockets\/puma.sock<\/code> adresindeki Puma soketine ba\u011flanmas\u0131n\u0131 sa\u011flar. Capistrano, her da\u011f\u0131t\u0131mda current<\/code> sembolik ba\u011flant\u0131s\u0131n\u0131 g\u00fcncelledi\u011fi i\u00e7in Nginx’in bu yap\u0131land\u0131rmay\u0131 yeniden y\u00fcklemesi gerekir. capistrano-nginx<\/code> gem’i bunu otomatik olarak yapar.<\/p>\nDa\u011f\u0131t\u0131m Komutlar\u0131 ve \u0130\u015fleyi\u015fi<\/h4>\n
cap production deploy<\/code><\/p>\ndeploy:starting<\/code><\/strong>: Da\u011f\u0131t\u0131m ba\u015flamadan \u00f6nce baz\u0131 haz\u0131rl\u0131klar yap\u0131l\u0131r.
\n2. git:check<\/code><\/strong>: Git deposuna eri\u015fimin kontrol\u00fc.
\n3. git:clone<\/code><\/strong>: Yeni kodun sunucudaki releases<\/code> dizinine kopyalanmas\u0131. Her da\u011f\u0131t\u0131m i\u00e7in yeni bir dizin olu\u015fturulur (\u00f6rn. 20231027123456<\/code>).
\n4. deploy:updating<\/code><\/strong>: linked_files<\/code> ve linked_dirs<\/code> i\u00e7in sembolik ba\u011flant\u0131lar olu\u015fturulur.
\n5. bundle:install<\/code><\/strong>: Uygulama ba\u011f\u0131ml\u0131l\u0131klar\u0131 (gem’ler) vendor\/bundle<\/code> dizinine y\u00fcklenir.
\n6. rails:db_migrate<\/code><\/strong>: Veritaban\u0131 migrasyonlar\u0131 \u00e7al\u0131\u015ft\u0131r\u0131l\u0131r. Bu ad\u0131m kritik \u00f6nem ta\u015f\u0131r.<\/strong> S\u0131f\u0131r kesinti i\u00e7in migrasyonlar\u0131n non-blocking ve geriye d\u00f6n\u00fck uyumlu olmas\u0131 gerekir. Yeni kodun gerektirdi\u011fi migrasyonlar, eski kodun hala \u00e7al\u0131\u015ft\u0131\u011f\u0131 bir ortamda \u00e7al\u0131\u015f\u0131r.
\n7. rails:assets_precompile<\/code><\/strong>: Frontend assetleri (CSS, JavaScript) derlenir.
\n8. deploy:publishing<\/code><\/strong>: current<\/code> sembolik ba\u011flant\u0131s\u0131, yeni da\u011f\u0131t\u0131lan s\u00fcr\u00fcm dizinini i\u015faret edecek \u015fekilde g\u00fcncellenir. Bu, Nginx’in art\u0131k yeni kodun public<\/code> dizinindeki statik dosyalara eri\u015febilece\u011fi anlam\u0131na gelir.
\n9. puma:phased-restart<\/code><\/strong>: Bu, s\u0131f\u0131r kesinti da\u011f\u0131t\u0131m\u0131n\u0131n kalbidir. Capistrano, Puma master i\u015flemine SIGUSR2<\/code> sinyali g\u00f6nderir.
\n * Puma master, yeni kodla yeni worker i\u015flemleri ba\u015flat\u0131r.
\n * Yeni worker’lar haz\u0131r oldu\u011funda, master eski worker’lara SIGTERM<\/code> g\u00f6nderir.
\n * Eski worker’lar, mevcut isteklerini tamamlar ve kapan\u0131r.
\n * Nginx, otomatik olarak yeni Puma soketine y\u00f6nlendirilmeye ba\u015flar.
\n10. nginx:restart<\/code> (E\u011fer capistrano-nginx<\/code> kullan\u0131l\u0131yorsa)<\/strong>: Nginx yap\u0131land\u0131rmas\u0131n\u0131 yeniden y\u00fckleyerek yeni current<\/code> sembolik ba\u011flant\u0131s\u0131n\u0131 ve soket yolunu tan\u0131mas\u0131n\u0131 sa\u011flar.
\n11. deploy:cleanup<\/code><\/strong>: Eski s\u00fcr\u00fcmlerden belirlenen say\u0131dan fazlas\u0131 (\u00f6rn. keep_releases<\/code> ile 5) silinir.
\n12. deploy:finished<\/code><\/strong>: Da\u011f\u0131t\u0131m tamamlan\u0131r.<\/p>\nVeritaban\u0131 Migrasyonlar\u0131 ve S\u0131f\u0131r Kesinti<\/h3>\n
ALTER TABLE<\/code> i\u015flemleri (\u00f6rne\u011fin, b\u00fcy\u00fck bir tabloya yeni bir s\u00fctun ekleme), tabloyu kilitler ve uygulaman\u0131n o tabloya eri\u015fimini engeller. Bu, da\u011f\u0131t\u0131m s\u0131ras\u0131nda kesintiye yol a\u00e7ar. Bu t\u00fcr migrasyonlar i\u00e7in pt-online-schema-change<\/code> (MySQL i\u00e7in) veya strong_migrations<\/code> gem’i gibi ara\u00e7lar kullan\u0131lmal\u0131d\u0131r. Bu ara\u00e7lar, tabloyu kilitlemeden \u015fema de\u011fi\u015fiklikleri yapman\u0131za olanak tan\u0131r.
\n* Geriye D\u00f6n\u00fck Uyumluluk<\/strong>: Yeni kodunuzun gerektirdi\u011fi bir veritaban\u0131 \u015fema de\u011fi\u015fikli\u011fi yapt\u0131\u011f\u0131n\u0131zda, bu de\u011fi\u015fikli\u011fin eski kodunuz taraf\u0131ndan da sorunsuz bir \u015fekilde i\u015flenebilir olmas\u0131 gerekir. \u00d6rne\u011fin, bir s\u00fctunu yeniden adland\u0131r\u0131yorsan\u0131z, \u00f6nce yeni s\u00fctunu eklemeli, veriyi ta\u015f\u0131mal\u0131, her iki s\u00fctunu da okuyup yazacak \u015fekilde kodu da\u011f\u0131tmal\u0131, ard\u0131ndan eski s\u00fctunu kald\u0131racak migrasyonu ayr\u0131 bir da\u011f\u0131t\u0131mda yapmal\u0131s\u0131n\u0131z. Bu, “iki a\u015famal\u0131 migrasyon” olarak bilinir.
\n* release<\/code> ile Durum<\/strong>: Da\u011f\u0131t\u0131m\u0131n db_migrate<\/code> ad\u0131m\u0131, current<\/code> sembolik linki g\u00fcncellenmeden \u00f6nce \u00e7al\u0131\u015ft\u0131r\u0131l\u0131r. Bu, migrasyonlar\u0131n eski kodun hala aktif oldu\u011fu bir ortamda ger\u00e7ekle\u015fti\u011fi anlam\u0131na gelir. Bu nedenle, migrasyonlar asla eski kodu bozmamal\u0131d\u0131r.<\/p>\nPotansiyel Sorunlar ve \u00c7\u00f6z\u00fcmler<\/h3>\n
strong_migrations<\/code> gibi gem’ler veya online schema change ara\u00e7lar\u0131 kullanmak, migrasyonlar\u0131 k\u00fc\u00e7\u00fck ad\u0131mlara b\u00f6lmek ve geriye d\u00f6n\u00fck uyumlulu\u011fu sa\u011flamakt\u0131r.
\n* Yetersiz Kaynaklar (CPU, RAM)<\/strong>: Yeni Puma worker’lar\u0131 ba\u015flat\u0131l\u0131rken, sunucunun yeterli CPU ve RAM’e sahip olmas\u0131 gerekir. Aksi takdirde, yeni worker’lar ba\u015flamakta zorlanabilir veya mevcut worker’lar yava\u015flayabilir. Sunucu kaynaklar\u0131n\u0131 izlemek ve gerekti\u011finde \u00f6l\u00e7eklendirmek \u00f6nemlidir.
\n* Ba\u011f\u0131ml\u0131l\u0131k \u00c7ak\u0131\u015fmalar\u0131<\/strong>: Yeni kod, eski kodla uyumsuz gem s\u00fcr\u00fcmleri gerektirebilir. Gemfile.lock<\/code> dosyas\u0131n\u0131n do\u011fru bir \u015fekilde y\u00f6netildi\u011finden ve bundle install<\/code> ad\u0131m\u0131n\u0131n her da\u011f\u0131t\u0131mda do\u011fru bir \u015fekilde \u00e7al\u0131\u015ft\u0131\u011f\u0131ndan emin olun.
\n* Cache Invalidasyonu<\/strong>: Da\u011f\u0131t\u0131m sonras\u0131 \u00f6nbelleklerin (Redis, Memcached) do\u011fru bir \u015fekilde temizlenmesi veya g\u00fcncellenmesi gerekebilir. Aksi takdirde, kullan\u0131c\u0131lar eski verileri g\u00f6rmeye devam edebilir.
\n* Rollback Stratejileri<\/strong>: Bir da\u011f\u0131t\u0131m ba\u015far\u0131s\u0131z olursa veya hatalar i\u00e7erirse, h\u0131zl\u0131 bir \u015fekilde \u00f6nceki s\u00fcr\u00fcme geri d\u00f6nebilmek kritik \u00f6neme sahiptir. Capistrano’nun cap production deploy:revert<\/code> komutu, current<\/code> sembolik ba\u011flant\u0131s\u0131n\u0131 \u00f6nceki kararl\u0131 s\u00fcr\u00fcme \u00e7evirerek h\u0131zl\u0131 bir geri d\u00f6n\u00fc\u015f sa\u011flar. Ancak veritaban\u0131 migrasyonlar\u0131 geriye al\u0131n\u0131rken dikkatli olunmal\u0131d\u0131r. Geriye d\u00f6n\u00fck uyumlu migrasyonlar, rollback’i kolayla\u015ft\u0131r\u0131r.<\/p>\nAlternatif Yakla\u015f\u0131mlar ve Ara\u00e7lar<\/h3>\n
\n* Mevcut PaaS \u00c7\u00f6z\u00fcmleri (Heroku, AWS Elastic Beanstalk)<\/strong>: Bu platformlar, s\u0131f\u0131r kesinti da\u011f\u0131t\u0131m mekanizmalar\u0131n\u0131 genellikle kendi i\u00e7lerinde sunar. Geli\u015ftiriciler, altyap\u0131 y\u00f6netimiyle u\u011fra\u015fmak zorunda kalmadan kodlar\u0131n\u0131 da\u011f\u0131tabilirler.
\n* Blue\/Green Deployments<\/strong>: Bu stratejide, uygulaman\u0131n iki tamamen ayr\u0131, \u00f6zde\u015f ortam\u0131 (Blue ve Green) bulunur. Yeni s\u00fcr\u00fcm Green ortam\u0131na da\u011f\u0131t\u0131l\u0131r ve test edilir. Her \u015fey yolunda gitti\u011finde, trafik y\u00f6nlendiricisi (load balancer) Blue ortam\u0131ndan Green ortam\u0131na ge\u00e7irilir. Eski Blue ortam\u0131, olas\u0131 bir geri d\u00f6n\u00fc\u015f i\u00e7in haz\u0131r bekletilir veya kapat\u0131l\u0131r.
\n* Canary Deployments<\/strong>: Yeni s\u00fcr\u00fcm, ba\u015flang\u0131\u00e7ta kullan\u0131c\u0131lar\u0131n k\u00fc\u00e7\u00fck bir y\u00fczdesine (kanarya grubu) da\u011f\u0131t\u0131l\u0131r. Bu kullan\u0131c\u0131 grubundan gelen geri bildirimler izlenir. Herhangi bir sorun olmazsa, yeni s\u00fcr\u00fcm kademeli olarak daha fazla kullan\u0131c\u0131ya yay\u0131l\u0131r.<\/p>\nSonu\u00e7<\/h3>\n