Odnoklassniki shine mafi girman mai amfani da Apache Cassandra akan RuNet kuma ɗayan mafi girma a duniya. Mun fara amfani da Cassandra a cikin 2010 don adana ƙimar hoto, kuma yanzu Cassandra yana sarrafa petabytes na bayanai akan dubban nodes, a zahiri, har ma mun haɓaka namu. .
A ranar 12 ga Satumba a ofishinmu na St. Petersburg za mu rike . Babban mai magana na taron zai kasance babban injiniyan Odnoklassniki Oleg Anastasyev. Oleg kwararre ne a fagen rarrabawa da tsarin juriya ga kuskure; yana aiki tare da Cassandra fiye da shekaru 10 kuma akai-akai. .
A jajibirin taron, mun yi magana da Oleg game da rashin haƙuri na tsarin da aka rarraba tare da Cassandra, ya tambayi abin da zai yi magana game da taron kuma dalilin da yasa ya cancanci halartar wannan taron.
Oleg ya fara aikinsa na shirye-shirye a 1995. Ya ɓullo da software a banki, sadarwa, da sufuri. Ya kasance yana aiki a matsayin babban mai haɓakawa a Odnoklassniki tun 2007 akan ƙungiyar dandamali. Ayyukansa sun haɗa da haɓaka gine-ginen gine-gine da mafita don tsarin kayan aiki mai girma, manyan ɗakunan ajiya na bayanai, da magance matsalolin aikin tashar tashar da kuma dogara. Yana kuma horar da masu haɓakawa a cikin kamfanin.
- Oleg, hello! A watan Mayu ya faru , sadaukarwa ga Apache Cassandra, mahalarta sun ce tattaunawar ta ci gaba har zuwa dare, don Allah ku gaya mani, menene ra'ayin ku game da haduwar farko?
Masu haɓakawa da nau'o'i daban-daban daga kamfanoni daban-daban sun zo tare da ciwon kansu, hanyoyin da ba zato ba tsammani ga matsalolin da labarun ban mamaki. Mun yi nasarar gudanar da mafi yawan taron a cikin tsarin tattaunawa, amma akwai tattaunawa da yawa wanda kawai muka iya tabo kashi uku na batutuwan da aka tsara. Mun ba da hankali sosai ga yadda da abin da muke saka idanu ta amfani da misalin ayyukan samar da mu na ainihi.
Ina sha'awar kuma ina son shi sosai.
- Dangane da sanarwar. za a duƙufa gaba ɗaya don haƙuri ga kuskure, me yasa kuka zaɓi wannan batu?
Cassandra wani tsarin rarrabawa ne na yau da kullun tare da babban adadin ayyuka sama da yin hidimar buƙatun mai amfani kai tsaye: tsegumi, gano gazawa, yaɗa canje-canjen makirci, faɗaɗa tari / raguwa, anti-entropy, madadin ajiya da dawo da, da sauransu. Kamar yadda yake a cikin kowane tsarin da aka rarraba, yayin da adadin kayan aiki ya karu, yuwuwar gazawar yana ƙaruwa, don haka aiki na gungu na samar da Cassandra yana buƙatar zurfin fahimtar tsarin sa don hasashen hali idan akwai gazawa da ayyukan ma'aikata. Bayan amfani da Cassandra shekaru da yawa, mu , wanda muna shirye mu raba, kuma muna so mu tattauna yadda abokan aiki a cikin shagon ke magance matsalolin da aka saba.
— Idan ya zo ga Cassandra, me kuke nufi da haƙurin kuskure?
Da farko, ba shakka, ikon tsarin don tsira da raunin kayan aiki na yau da kullun: asarar injuna, fayafai ko haɗin cibiyar sadarwa tare da nodes / cibiyoyin bayanai. Amma batun da kansa ya fi girma kuma musamman ya haɗa da farfadowa daga gazawar, ciki har da gazawar da ba a shirya wa mutane ba, misali, kurakurai na ma'aikata.
- Shin za ku iya ba da misalin mafi yawan ɗorawa da manyan bayanai?
Ɗaya daga cikin manyan gungu na mu shine tarin kyauta: fiye da nodes 200 da daruruwan tarin tarin fuka na bayanai. Amma ba shine mafi lodi ba, tun da yake an rufe shi da ma'ajin da aka rarraba. Rukunin mu mafi yawan aiki suna ɗaukar dubun dubatar RPS don rubutu da dubunnan RPS don karatu.
- Kai! Sau nawa abu yakan karye?
! Gabaɗaya, muna da sabobin fiye da dubu 6, kuma kowane mako ana maye gurbin sabar sabar da dozin da yawa (ba tare da la'akari da daidaitattun hanyoyin haɓakawa da haɓaka na'urar ba). Ga kowane nau'in gazawar, akwai cikakkun bayanai game da abin da za a yi kuma a cikin wane tsari, komai yana sarrafa kansa a duk lokacin da zai yiwu, don haka gazawa na yau da kullun kuma a cikin 99% na lokuta suna faruwa ba tare da lura da masu amfani ba.
- Yaya kuke magance irin wannan ƙi?
Tun daga farkon aikin Cassandra da abubuwan da suka faru na farko, mun yi aiki a kan hanyoyin don adanawa da dawo da su, gina hanyoyin jigilar kayayyaki waɗanda ke la'akari da yanayin gungu na Cassandra kuma, alal misali, ba mu yarda a sake kunna nodes ba. idan asarar bayanai zai yiwu. Muna shirin yin magana game da wannan duka a taron.
- Kamar yadda ka ce, babu cikakken ingantaccen tsarin. Wadanne nau'ikan gazawar kuke shirya don kuma kuna iya jurewa?
Idan muka yi magana game da shigar mu na gungu na Cassandra, masu amfani ba za su lura da komai ba idan muka rasa injuna da yawa a cikin DC ɗaya ko duka DC guda ɗaya (wannan ya faru). Tare da karuwa a yawan DCs, muna tunanin fara tabbatar da aiki a yayin da gazawar DC guda biyu.
- Me kuke tsammani Cassandra ya gaza game da hakuri da kuskure?
Cassandra, kamar sauran shagunan NoSQL na farko, na buƙatar zurfin fahimtar tsarin sa na ciki da tafiyar matakai masu ƙarfi da ke faruwa. Zan ce ba shi da sauƙi, tsinkaya da kuma lura. Amma zai zama abin sha'awa don jin ra'ayoyin sauran mahalarta taron!
Oleg, na gode sosai don ɗaukar lokaci don amsa tambayoyin!
Muna jiran duk wanda yake so ya yi magana da masana a fagen aiki Apache Cassandra a taron ranar 12 ga Satumba a ofishinmu na St. Petersburg.
Ku zo, zai zama mai ban sha'awa!
source: www.habr.com