Yadda muka tattara bayanai kan kamfen talla daga rukunin yanar gizo (hanyar ƙaya zuwa samfurin)

Da alama filin tallan kan layi ya kamata ya zama ci gaba da fasaha da sarrafa kansa gwargwadon yiwuwa. Tabbas, saboda irin waɗannan ƙattai da masana a fagen su kamar Yandex, Mail.Ru, Google da Facebook suna aiki a can. Amma, kamar yadda ya juya, babu iyaka ga kamala kuma koyaushe akwai wani abu don sarrafa kansa.

Source

Ƙungiyar Sadarwa Dentsu Aegis Network Russia shine mafi girman dan wasa a kasuwar tallan dijital kuma yana saka hannun jari sosai a fasaha, yana ƙoƙarin haɓakawa da sarrafa ayyukan kasuwancin sa. Ɗaya daga cikin matsalolin da ba a warware ba na kasuwannin tallace-tallace na kan layi shine aikin tattara kididdiga kan kamfen ɗin talla daga dandamali na Intanet daban-daban. Maganin wannan matsala a ƙarshe ya haifar da ƙirƙirar samfur D1.Digital (karanta kamar DiVan), ci gaban da muke so muyi magana akai.

Me ya sa?

1. A lokacin da aka fara aikin, babu wani samfurin da aka shirya a kasuwa wanda ya warware matsalar ta atomatik tattara tarin ƙididdiga akan yakin talla. Wannan yana nufin babu wanda sai kanmu da zai biya mana bukatunmu.

Ayyuka irin su Improvado, Roistat, Supermetrics, SegmentStream suna ba da haɗin kai tare da dandamali, cibiyoyin sadarwar jama'a da Google Analitycs, kuma suna ba da damar gina dashboards na nazari don ingantaccen bincike da sarrafa yakin talla. Kafin mu fara haɓaka samfuranmu, mun yi ƙoƙarin amfani da wasu daga cikin waɗannan tsarin don tattara bayanai daga shafuka, amma, rashin alheri, sun kasa magance matsalolinmu.

Babban matsalar ita ce samfuran da aka gwada sun dogara ne akan tushen bayanai, suna nuna kididdigar wuri ta wurin, kuma ba su ba da ikon tattara kididdiga kan yakin talla ba. Wannan tsarin bai ba mu damar ganin kididdiga daga shafuka daban-daban a wuri guda da kuma nazarin yanayin yakin gaba daya ba.

Wani abu kuma shi ne cewa a farkon matakan da aka yi amfani da su a kasuwannin yammacin Turai kuma ba su goyi bayan haɗin kai tare da shafukan Rasha ba. Kuma ga waɗancan rukunin yanar gizon da aka aiwatar da haɗin kai, duk matakan da ake buƙata ba koyaushe ana zazzage su tare da cikakkun bayanai ba, kuma haɗin kai ba koyaushe ya dace ba kuma a bayyane, musamman lokacin da ya zama dole don samun wani abu wanda ba a cikin tsarin tsarin ba.
Gabaɗaya, mun yanke shawarar kada mu daidaita da samfuran ɓangare na uku, amma mun fara haɓaka namu ...

2. Kasuwar talla ta yanar gizo tana karuwa daga shekara zuwa shekara, kuma a cikin 2018, ta fuskar kasafin kudin talla, ta mamaye kasuwar talla ta TV mafi girma a al'ada. Don haka akwai ma'auni.

3. Ba kamar kasuwar tallace-tallace ta TV ba, inda tallace-tallacen tallace-tallace ya mamaye, akwai mutane da yawa masu mallakin tallace-tallace na tallace-tallace masu girma dabam suna aiki a Intanet tare da asusun talla na kansu. Tun da yakin talla, a matsayin mai mulkin, yana gudana a kan shafuka da yawa a lokaci ɗaya, don fahimtar yanayin yakin tallace-tallace, ya zama dole a tattara rahotanni daga duk shafukan yanar gizo kuma a haɗa su cikin babban rahoto guda ɗaya wanda zai nuna cikakken hoto. Wannan yana nufin akwai yuwuwar ingantawa.

4. Ya zama kamar a gare mu cewa masu kayan talla a Intanet sun riga sun sami kayan aikin tattara ƙididdiga da nuna su a cikin asusun talla, kuma za su iya samar da API don wannan bayanan. Wannan yana nufin cewa a zahiri yana yiwuwa a aiwatar da shi. Bari mu ce nan da nan ya zama ba mai sauƙi ba.

Gabaɗaya, duk abubuwan da ake buƙata don aiwatar da aikin sun bayyana a gare mu, kuma mun gudu don ganin aikin a rayuwa ...

Babban shiri

Da farko, mun kafa hangen nesa na ingantaccen tsari:

• Ya kamata a loda kamfen tallace-tallace daga tsarin kamfanoni na 1C ta atomatik a ciki tare da sunayensu, lokutansu, kasafin kuɗi da wuraren sanyawa akan dandamali daban-daban.
• Ga kowane wuri a cikin yakin talla, duk yuwuwar ƙididdiga ya kamata a zazzage ta kai tsaye daga rukunin yanar gizon da ke faruwa, kamar adadin abubuwan gani, dannawa, ra'ayoyi, da sauransu.
• Ana bin wasu kamfen ɗin talla ta amfani da sa ido na ɓangare na uku ta abin da ake kira tsarin talla kamar Adriver, Weborama, DCM, da sauransu. Har ila yau, akwai mitar Intanet na masana'antu a Rasha - kamfanin Mediascope. Bisa ga shirin mu, bayanai daga masu zaman kansu da kuma sa ido na masana'antu ya kamata kuma a ɗora su ta atomatik cikin kamfen ɗin talla.
• Yawancin kamfen ɗin talla akan Intanet suna nufin wasu ayyukan da aka yi niyya (saya, kira, rajista don gwajin gwaji, da sauransu), waɗanda ake bin diddigin su ta amfani da Google Analytics, da ƙididdiga waɗanda suma suna da mahimmanci don fahimtar matsayin yaƙin neman zaɓe kuma ya kamata a ɗora a cikin kayan aikin mu.

Pancake na farko yana da dunƙulewa

Ganin sadaukarwarmu ga sassauƙan ƙa'idodin haɓaka software (agile, komai), mun yanke shawarar fara haɓaka MVP sannan mu matsa zuwa ga burin da aka yi niyya akai-akai.
Mun yanke shawarar gina MVP bisa samfurin mu DANBo (Dentsu Aegis Network Board), wanda shine aikace-aikacen yanar gizo tare da cikakkun bayanai game da yakin tallan abokan cinikinmu.

Don MVP, an sauƙaƙe aikin kamar yadda zai yiwu dangane da aiwatarwa. Mun zaɓi taƙaitaccen jerin dandamali don haɗin kai. Waɗannan su ne manyan dandamali, irin su Yandex.Direct, Yandex.Display, RB.Mail, MyTarget, Adwords, DBM, VK, FB, da kuma babban tsarin tallan tallan tallace-tallace da Weborama.

Don samun damar ƙididdiga akan shafuka ta API, mun yi amfani da asusu ɗaya. Manajan ƙungiyar abokin ciniki wanda ke son yin amfani da tarin ƙididdiga ta atomatik akan yaƙin neman zaɓe dole ne ya fara ba da damar yin amfani da kamfen ɗin talla masu mahimmanci akan shafuka zuwa asusun dandamali.

Na gaba shine mai amfani da tsarin DANBo dole ne a loda fayil na wani tsari a cikin tsarin Excel, wanda ya ƙunshi duk bayanan game da sanyawa (kamfen tallace-tallace, dandamali, tsari, lokacin sanyawa, alamomin da aka tsara, kasafin kuɗi, da sauransu) da masu gano kamfen ɗin talla masu dacewa akan shafukan yanar gizo da masu lissafi a tsarin talla.

Ya duba, gaskiya, mai ban tsoro:

An adana bayanan da aka zazzage a cikin rumbun adana bayanai, sannan ayyuka daban-daban sun tattara abubuwan gano kamfen akan shafuka daga cikinsu kuma an zazzage kididdigar a kansu.

Ga kowane rukunin yanar gizon, an rubuta sabis ɗin windows daban, wanda sau ɗaya kowace rana yana shiga ƙarƙashin asusun sabis ɗaya a cikin API ɗin rukunin kuma an zazzage ƙididdiga don ƙayyadaddun ID na yakin neman zabe. Haka abin ya faru da tsarin talla.

An nuna bayanan da aka zazzage akan mahaɗin a cikin sigar ƙaramin dashboard na al'ada:

Ba zato ba tsammani a gare mu, MVP ya fara aiki kuma ya fara zazzage kididdiga na yanzu akan kamfen talla akan Intanet. Mun aiwatar da tsarin akan abokan ciniki da yawa, amma lokacin ƙoƙarin ƙima, mun ci karo da manyan matsaloli:

• Babban matsalar ita ce rikitarwar shirya bayanai don lodawa cikin tsarin. Hakanan, bayanan jeri dole ne a canza su zuwa tsayayyen tsari kafin lodawa. Ya zama dole a haɗa abubuwan gano mahaɗan daga shafuka daban-daban a cikin fayil ɗin zazzagewa. Mun fuskanci gaskiyar cewa yana da matukar wahala ga masu amfani waɗanda ba a horar da su ba don bayyana inda za su sami waɗannan abubuwan ganowa akan rukunin yanar gizon da kuma inda a cikin fayil ɗin suke buƙatar shigar da su. Idan aka yi la'akari da yawan ma'aikatan da ke cikin sassan da ke gudanar da yakin neman zabe a shafukan yanar gizo da kuma canji, wannan ya haifar da goyon baya mai yawa a gefenmu, wanda ba mu yi farin ciki da shi ba.
• Wata matsala kuma ita ce, ba duk hanyoyin talla ba ne ke da hanyoyin ba da damar yin kamfen ɗin talla zuwa wasu asusu. Amma ko da tsarin wakilai ya kasance, ba duk masu talla ba ne ke son ba da damar yin amfani da kamfen ɗin su zuwa asusun ɓangare na uku.
• Wani muhimmin al'amari shi ne fushin da ya taso a tsakanin masu amfani da gaskiyar cewa duk abubuwan da aka tsara da kuma cikakkun bayanai na sanyawa waɗanda suka riga sun shiga cikin tsarin lissafin mu na 1C, dole ne su sake shiga cikin DANBo.

Wannan ya ba mu ra'ayin cewa tushen tushen bayanai game da sanyawa ya kamata ya zama tsarinmu na 1C, wanda aka shigar da duk bayanan daidai kuma a kan lokaci (ma'anar a nan ita ce ana samar da daftari bisa bayanan 1C, don haka daidai shigar da bayanai cikin 1C. shine fifiko ga kowa da kowa KPI). Wannan shine yadda sabon tunanin tsarin ya fito...

Tunani

Abu na farko da muka yanke shawarar yi shine raba tsarin tattara kididdiga akan kamfen talla akan Intanet zuwa wani samfuri daban- D1.Digital.

A cikin sabon ra'ayi, mun yanke shawarar loda cikin D1.Digital bayanai kan kamfen tallace-tallace da sanyawa a cikin su daga 1C, sannan a fitar da kididdiga daga shafuka da tsarin AdServing zuwa waɗannan wuraren. Wannan ya kamata ya sauƙaƙe rayuwa ga masu amfani (kuma, kamar yadda aka saba, ƙara ƙarin aiki ga masu haɓakawa) da rage adadin tallafi.

Matsala ta farko da muka ci karo da ita ita ce yanayin kungiya kuma tana da alaƙa da gaskiyar cewa ba za mu iya samun maɓalli ko alamar da za mu iya kwatanta ƙungiyoyi daga tsarin daban-daban tare da kamfen da jeri daga 1C. Gaskiyar ita ce, tsari a cikin kamfaninmu an tsara shi ta hanyar da za a shigar da yakin talla a cikin tsarin daban-daban ta hanyar mutane daban-daban (masu tsara shirye-shiryen watsa labaru, sayayya, da dai sauransu).

Don magance wannan matsala, dole ne mu ƙirƙira maɓalli na musamman, DANBoID, wanda zai haɗa ƙungiyoyi a cikin tsari daban-daban tare, kuma ana iya gano shi cikin sauƙi da musamman a cikin abubuwan da aka zazzage. An ƙirƙiri wannan mai ganowa a cikin tsarin 1C na ciki don kowane wuri ɗaya kuma ana tura shi zuwa yaƙin neman zaɓe, wurare da ƙididdiga akan duk rukunin yanar gizo da kuma cikin duk tsarin AdServing. Aiwatar da al'adar sanya DANBoID a duk wuraren da aka ba da izini ya ɗauki ɗan lokaci, amma mun sami damar yin shi :)

Daga nan sai muka gano cewa ba dukkan shafuka ke da API don tattara kididdiga ta atomatik ba, har ma wadanda ke da API, ba ya mayar da duk bayanan da ake bukata.

A wannan mataki, mun yanke shawarar rage yawan jerin dandamali don haɗin kai kuma mu mai da hankali kan manyan dandamali waɗanda ke da hannu a cikin mafi yawan kamfen ɗin talla. Wannan jeri ya haɗa da duk manyan 'yan wasa a kasuwar talla (Google, Yandex, Mail.ru), hanyoyin sadarwar zamantakewa (VK, Facebook, Twitter), manyan AdServing da tsarin nazari (DCM, Adriver, Weborama, Google Analytics) da sauran dandamali.

Yawancin rukunin yanar gizon da muka zaɓa suna da API wanda ya ba da ma'aunin da muke buƙata. A cikin yanayin da babu API ko kuma ba ta ƙunshi bayanan da ake buƙata ba, mun yi amfani da rahotannin da aka aika yau da kullun zuwa imel ɗin ofishinmu don loda bayanai (a wasu tsarin yana yiwuwa a saita irin waɗannan rahotanni, a cikin wasu mun yarda da haɓaka irin waɗannan rahotanni). domin mu).

Lokacin da muke nazarin bayanai daga shafuka daban-daban, mun gano cewa matsayi na ƙungiyoyi ba ɗaya ba ne a cikin tsari daban-daban. Bugu da ƙari, ana buƙatar zazzage bayanai dalla-dalla daga tsarin daban-daban.

Don magance wannan matsalar, an ƙirƙiri tunanin SubDANBoID. Tunanin SubDANBoID abu ne mai sauƙi, muna yiwa babban ƙungiyar yaƙin neman zaɓe akan rukunin yanar gizon tare da DANBoID da aka samar, kuma muna loda duk rukunin gida tare da masu gano rukunin yanar gizo na musamman kuma muna samar da SubDANBoID bisa ga ka'idar DANBoID + mai gano matakin farko. mahaɗan gida + mai gano mahalli na biyu na gida +... Wannan tsarin ya ba mu damar haɗa kamfen ɗin talla a cikin tsarin daban-daban da zazzage ƙididdiga dalla-dalla akan su.

Har ila yau, dole ne mu magance matsalar samun damar yin kamfen a kan dandamali daban-daban. Kamar yadda muka rubuta a sama, tsarin ba da damar shiga yaƙin neman zaɓe zuwa wani asusun fasaha na daban ba koyaushe yake aiki ba. Don haka, dole ne mu haɓaka kayan aikin don ba da izini ta atomatik ta OAuth ta amfani da alamu da hanyoyin sabunta waɗannan alamun.

Daga baya a cikin labarin za mu yi ƙoƙari mu bayyana dalla-dalla dalla-dalla tsarin gine-ginen mafita da bayanan fasaha na aiwatarwa.

Magani gine-gine 1.0

Lokacin fara aiwatar da sabon samfurin, mun fahimci cewa nan da nan muna buƙatar samar da yiwuwar haɗa sabbin shafuka, don haka mun yanke shawarar bin hanyar gine-ginen microservice.

Lokacin zayyana gine-ginen, mun raba masu haɗin kai zuwa duk tsarin waje - 1C, dandamalin talla da tsarin talla - cikin sabis daban-daban.
Babban ra'ayin shi ne cewa duk masu haɗin yanar gizo suna da API iri ɗaya kuma su ne adaftan da ke kawo API na rukunin yanar gizon zuwa wurin da ya dace da mu.

A tsakiyar samfurin mu akwai aikace-aikacen yanar gizo, wanda shine monolith wanda aka tsara ta yadda za'a iya haɗa shi cikin sauƙi zuwa sabis. Wannan aikace-aikacen yana da alhakin sarrafa bayanan da aka zazzage, tattara ƙididdiga daga tsarin daban-daban da gabatar da su ga masu amfani da tsarin.

Don sadarwa tsakanin masu haɗawa da aikace-aikacen yanar gizo, dole ne mu ƙirƙiri ƙarin sabis, wanda muka kira Connector Proxy. Yana aiwatar da ayyukan Gano Sabis da Jadawalin Aiki. Wannan sabis ɗin yana gudanar da ayyukan tattara bayanai ga kowane mai haɗa kowane dare. Rubuta layin sabis ya kasance mafi sauƙi fiye da haɗa dillalin saƙo, kuma a gare mu yana da mahimmanci don samun sakamakon da sauri.

Don sauƙi da saurin haɓakawa, mun kuma yanke shawarar cewa duk ayyuka zasu zama APIs na Yanar Gizo. Wannan ya sa ya yiwu a hanzarta tattara hujja-na-ra'ayi da kuma tabbatar da cewa dukan zane yana aiki.

Wani aiki na daban, mai rikitarwa shine saita damar tattara bayanai daga asusu daban-daban, waɗanda, kamar yadda muka yanke shawara, yakamata masu amfani su aiwatar ta hanyar haɗin yanar gizo. Ya ƙunshi matakai daban-daban guda biyu: na farko, mai amfani yana ƙara alama don shiga asusun ta hanyar OAuth, sannan ya tsara tarin bayanai don abokin ciniki daga takamaiman asusu. Samun alamar ta hanyar OAuth ya zama dole saboda, kamar yadda muka riga muka rubuta, ba koyaushe zai yiwu a ba da damar shiga asusun da ake so akan rukunin yanar gizon ba.

Don ƙirƙirar tsarin duniya don zaɓar asusu daga shafuka, dole ne mu ƙara wata hanya zuwa API masu haɗawa da ke dawo da Tsarin JSON, wanda aka sanya shi cikin tsari ta amfani da gyaggyaran ɓangaren JSONEditor. Ta wannan hanyar, masu amfani sun sami damar zaɓar asusun da za a sauke bayanai daga ciki.

Don biyan iyakokin buƙatun da ke wanzu akan shafuka, muna haɗa buƙatun don saiti a cikin alama ɗaya, amma muna iya aiwatar da alamu daban-daban a layi daya.

Mun zaɓi MongoDB azaman ajiya don bayanan da aka ɗora don duka aikace-aikacen yanar gizo da masu haɗawa, wanda ya ba mu damar damuwa da yawa game da tsarin bayanai a farkon matakan haɓakawa, lokacin da samfurin abu na aikace-aikacen ya canza kowace rana.

Ba da daɗewa ba muka gano cewa ba duk bayanai sun dace da MongoDB ba kuma, alal misali, ya fi dacewa don adana ƙididdiga na yau da kullun a cikin bayanan alaƙa. Don haka, ga masu haɗin yanar gizo waɗanda tsarin bayanan su ya fi dacewa da bayanan alaƙa, mun fara amfani da PostgreSQL ko MS SQL Server azaman ajiya.

Zaɓaɓɓen gine-gine da fasaha sun ba mu damar ginawa da ƙaddamar da samfurin D1.Digital da sauri. Sama da shekaru biyu na haɓaka samfura, mun haɓaka masu haɗin 23 zuwa rukunin yanar gizon, sun sami gogewa mai ƙima da aiki tare da APIs na ɓangare na uku, sun koyi don guje wa ɓarna na shafuka daban-daban, waɗanda kowannensu yana da nasu, ya ba da gudummawa ga haɓaka API na akalla 3. shafukan yanar gizo, bayanan da aka sauke ta atomatik akan kusan yakin 15 da kuma fiye da wurare 000, sun tattara ra'ayoyin masu amfani da yawa game da aikin samfurin kuma sun sami damar canza babban tsari na samfurin sau da yawa, bisa ga wannan ra'ayi.

Magani gine-gine 2.0

Shekaru biyu ke nan da fara ci gaba D1.Digital. Ƙaruwa akai-akai a kan tsarin da kuma fitowar sababbin hanyoyin bayanai a hankali sun bayyana matsaloli a cikin tsarin gine-ginen da ake ciki.

Matsala ta farko tana da alaƙa da adadin bayanan da aka sauke daga rukunin yanar gizon. Mun fuskanci gaskiyar cewa tattarawa da sabunta duk mahimman bayanai daga manyan shafuka sun fara ɗaukar lokaci mai yawa. Misali, tattara bayanai daga tsarin tallan AdRiver, wanda da shi muke bin kididdigar mafi yawan wurare, yana ɗaukar awanni 12.

Don magance wannan matsala, mun fara amfani da kowane nau'in rahotanni don zazzage bayanai daga shafuka, muna ƙoƙarin haɓaka API ɗinsu tare da rukunin yanar gizon ta yadda saurin aikinsa ya dace da bukatunmu, da daidaita zazzagewar bayanai gwargwadon iko.

Wata matsala kuma ta shafi sarrafa bayanan da aka sauke. Yanzu, lokacin da sabbin kididdigar wurin zama ya zo, an ƙaddamar da wani tsari mai matakai da yawa na sake ƙididdige ma'auni, wanda ya haɗa da lodin ɗanyen bayanai, ƙididdige ma'auni ga kowane rukunin yanar gizo, kwatanta bayanai daga tushe daban-daban da juna, da ƙididdige ma'auni na taƙaitaccen kamfen. Wannan yana haifar da kaya mai yawa akan aikace-aikacen gidan yanar gizon da ke yin duk lissafin. Sau da yawa, yayin aikin sake ƙididdigewa, aikace-aikacen ya cinye duk ƙwaƙwalwar ajiya akan uwar garke, game da 10-15 GB, wanda ya fi tasiri ga aikin masu amfani da tsarin.

Matsalolin da aka gano da kuma tsare-tsare masu ban sha'awa don ci gaba da haɓaka samfurin sun kai mu ga buƙatar sake yin la'akari da gine-ginen aikace-aikacen.

Mun fara da connectors.
Mun lura cewa duk masu haɗawa suna aiki bisa ga samfurin iri ɗaya, don haka mun gina tsarin bututun mai wanda don ƙirƙirar haɗin kai kawai dole ne ku tsara dabaru na matakan, sauran sun kasance na duniya. Idan wasu masu haɗin suna buƙatar haɓakawa, to nan da nan za mu canza shi zuwa sabon tsarin lokaci guda yayin da ake haɓaka haɗin haɗin.

A lokaci guda, mun fara tura masu haɗin kai zuwa Docker da Kubernetes.
Mun shirya tafiya zuwa Kubernetes na dogon lokaci, gwaji tare da saitunan CI / CD, amma fara motsawa kawai lokacin da mai haɗawa ɗaya, saboda kuskure, ya fara cinye fiye da 20 GB na ƙwaƙwalwar ajiya akan uwar garke, kusan kashe sauran matakai. . Yayin binciken, an matsar da mai haɗin zuwa gungu na Kubernetes, inda a ƙarshe ya kasance, ko da bayan an gyara kuskuren.

Da sauri mun gane cewa Kubernetes ya dace, kuma a cikin watanni shida mun tura masu haɗin 7 da masu haɗawa da Proxy, waɗanda ke cinye mafi yawan albarkatu, zuwa gungu na samarwa.

Bayan masu haɗin, mun yanke shawarar canza tsarin gine-ginen sauran aikace-aikacen.
Babban matsalar ita ce data ke fitowa daga connectors zuwa proxies a cikin manyan batches, sannan a buga DANBoID kuma a aika zuwa aikace-aikacen yanar gizo na tsakiya don sarrafawa. Saboda yawan ƙididdigar ƙididdiga, akwai babban nauyi akan aikace-aikacen.

Hakanan ya kasance da wahala sosai don saka idanu kan matsayin ayyukan tattara bayanai na mutum ɗaya da ba da rahoton kurakuran da ke faruwa a cikin masu haɗin yanar gizo zuwa aikace-aikacen yanar gizo ta tsakiya domin masu amfani su ga abin da ke faruwa da kuma dalilin da yasa ba a tattara bayanai ba.

Don magance waɗannan matsalolin, mun haɓaka gine-gine 2.0.

Babban bambanci tsakanin sabon sigar gine-ginen shine cewa maimakon API ɗin Yanar Gizo, muna amfani da RabbitMQ da ɗakin karatu na MassTransit don musayar saƙonni tsakanin sabis. Don yin wannan, dole ne mu kusan sake rubuta Proxy Connectors gaba ɗaya, mai da shi Hub Haɗin. An canza sunan saboda babban aikin sabis ɗin baya cikin tura buƙatun zuwa masu haɗawa da baya, amma wajen sarrafa tarin ma'auni daga masu haɗawa.

Daga aikace-aikacen yanar gizo na tsakiya, mun raba bayanai game da jeri da ƙididdiga daga shafuka zuwa ayyuka daban-daban, wanda ya sa ya yiwu a kawar da ƙididdiga marasa mahimmanci kuma adana kawai an riga an ƙididdigewa da tara ƙididdiga a matakin jeri. Mun kuma sake rubutawa da haɓaka dabaru don ƙididdige ƙididdiga na asali dangane da ɗanyen bayanai.

A lokaci guda, muna ƙaura duk ayyuka da aikace-aikace zuwa Docker da Kubernetes don sauƙaƙa mafita don daidaitawa kuma mafi dacewa don sarrafawa.

Ina muke yanzu

Tabbacin-ra'ayin gine-gine 2.0 samfur D1.Digital shirye da aiki a cikin yanayin gwaji tare da iyakataccen saiti na masu haɗawa. Abin da kawai ya rage don yin shi ne sake rubuta wasu masu haɗin 20 zuwa sabon dandamali, gwada cewa an ɗora bayanan daidai kuma an ƙididdige duk ma'auni daidai, da kuma fitar da dukan zane a cikin samarwa.

A zahiri, wannan tsari zai faru a hankali kuma dole ne mu bar dacewa da baya tare da tsoffin APIs don kiyaye komai yana aiki.

Shirye-shiryen mu na gaggawa sun haɗa da haɓaka sababbin masu haɗin kai, haɗin kai tare da sababbin tsarin da ƙara ƙarin ma'auni zuwa saitin bayanan da aka sauke daga shafukan da aka haɗa da tsarin tallace-tallace.

Hakanan muna shirin canja wurin duk aikace-aikacen, gami da babban aikace-aikacen gidan yanar gizo, zuwa Docker da Kubernetes. Haɗe tare da sabon gine-gine, wannan zai sauƙaƙe ƙaddamarwa, kulawa da sarrafa albarkatun da aka cinye.

Wani ra'ayi shine a gwada zaɓin bayanan adana bayanai don adana ƙididdiga, wanda a halin yanzu ana adana shi a MongoDB. Mun riga mun canja wurin sabbin masu haɗawa da yawa zuwa bayanan bayanan SQL, amma akwai bambancin kusan ba a san shi ba, kuma ga ƙididdigar ƙididdiga ta rana, wanda za'a iya nema na wani lokaci na sabani, riba na iya zama mai tsanani.

Gabaɗaya, tsare-tsaren suna da girma, bari mu ci gaba :)

Marubutan labarin R&D Dentsu Aegis Network Russia: Georgy Ostapenko (shmiiga), Mikhail Kotsik (hitexx)

source: www.habr.com