{"id":93878,"date":"2020-09-10T19:42:02","date_gmt":"2020-09-10T17:42:02","guid":{"rendered":"https:\/\/prohoster.info\/blog\/administrirovanie\/kak-vpisat-svobodnuyu-postgresql-v-surovoe-enterprise-okruzhenie"},"modified":"2020-09-10T19:42:02","modified_gmt":"2020-09-10T17:42:02","slug":"kak-vpisat-svobodnuyu-postgresql-v-surovoe-enterprise-okruzhenie","status":"publish","type":"post","link":"https:\/\/prohoster.info\/cs\/blog\/administrirovanie\/kak-vpisat-svobodnuyu-postgresql-v-surovoe-enterprise-okruzhenie","title":{"rendered":"Jak integrovat 'volnou' PostgreSQL do p\u0159\u00edsn\u00e9ho podnikov\u00e9ho prost\u0159ed\u00ed","gt_translate_keys":[{"key":"rendered","format":"text"}]},"content":{"rendered":"

Mnoz\u00ed jsou obezn\u00e1meni s datab\u00e1zov\u00fdm syst\u00e9mem PostgreSQL, kter\u00fd se osv\u011bd\u010dil na men\u0161\u00edch instalac\u00edch. Nicm\u00e9n\u011b trend p\u0159echodu na Open Source je st\u00e1le v\u00fdrazn\u011bj\u0161\u00ed, a to i ve vztahu k velk\u00fdm firm\u00e1m a podnikov\u00fdch po\u017eadavk\u016fm. V tomto \u010dl\u00e1nku se pod\u00edv\u00e1me na to, jak integrovat Postgres do podnikov\u00e9ho prost\u0159ed\u00ed, a pod\u011bl\u00edme se o zku\u0161enosti se vytvo\u0159en\u00edm syst\u00e9mu z\u00e1lohov\u00e1n\u00ed (ZK) pro tuto datab\u00e1zi na p\u0159\u00edkladu z\u00e1lohovac\u00edho syst\u00e9mu Commvault. <\/p>\n

\"Jak
\nPostgreSQL ji\u017e prok\u00e1zala svou spolehlivost \u2014 datab\u00e1zov\u00fd syst\u00e9m skv\u011ble funguje, pou\u017e\u00edvaj\u00ed ho modern\u00ed digit\u00e1ln\u00ed podniky jako Alibaba a TripAdvisor, a absence licen\u010dn\u00edch poplatk\u016f ji \u010din\u00ed l\u00e1kavou alternativou k takov\u00fdm gigant\u016fm jako MS SQL nebo Oracle DB. Jakmile v\u0161ak za\u010dneme p\u0159em\u00fd\u0161let o PostgreSQL v kontextu podnik\u016f, okam\u017eit\u011b nar\u00e1\u017e\u00edme na p\u0159\u00edsn\u00e9 po\u017eadavky: \"Co t\u0159eba odolnost konfigurace? Odolnost proti katastrof\u00e1m? Kde je komplexn\u00ed monitorov\u00e1n\u00ed? Jak\u00e9 je automatizovan\u00e9 z\u00e1lohov\u00e1n\u00ed? Jak\u00e9 je pou\u017eit\u00ed p\u00e1skov\u00fdch knihoven jak p\u0159\u00edmo, tak jako sekund\u00e1rn\u00ed \u00falo\u017ei\u0161t\u011b?\"
\n<\/a><\/noindex>
\n\"Jak
\nNa jedn\u00e9 stran\u011b PostgreSQL nem\u00e1 vestav\u011bn\u00e9 n\u00e1stroje pro z\u00e1lohov\u00e1n\u00ed jako \"dosp\u011bl\u00e9\" datab\u00e1ze typu RMAN u Oracle DB nebo SAP Database Backup. Na druh\u00e9 stran\u011b, dodavatel\u00e9 korpor\u00e1tn\u00edch z\u00e1lohovac\u00edch syst\u00e9m\u016f (Veeam, Veritas, Commvault) i kdy\u017e podporuj\u00ed PostgreSQL, ve skute\u010dnosti pracuj\u00ed pouze s ur\u010ditou (obvykle standalone) konfigurac\u00ed a s r\u016fzn\u00fdmi omezen\u00edmi. <\/p>\n

Syst\u00e9my z\u00e1lohov\u00e1n\u00ed navr\u017een\u00e9 speci\u00e1ln\u011b pro PostgreSQL, jako jsou Barman, Wal-g, pg_probackup, jsou velmi popul\u00e1rn\u00ed v mal\u00fdch instalac\u00edch datab\u00e1ze PostgreSQL nebo tam, kde nejsou pot\u0159eba t\u011b\u017ek\u00e9 z\u00e1lohy dal\u0161\u00edch prvk\u016f IT prost\u0159ed\u00ed. Nap\u0159\u00edklad, krom\u011b PostgreSQL m\u016f\u017ee b\u00fdt v infrastruktu\u0159e fyzick\u00fd a virtu\u00e1ln\u00ed OpenShift, Oracle, MariaDB, Cassandra atd. V\u0161e by m\u011blo b\u00fdt z\u00e1lohov\u00e1no spole\u010dn\u00fdm n\u00e1strojem. Pou\u017eit\u00ed samostatn\u00e9ho \u0159e\u0161en\u00ed pouze pro PostgreSQL je ne\u0161\u0165astn\u00e1 my\u0161lenka: data budou n\u011bkde kop\u00edrov\u00e1na na disk a pak je t\u0159eba je p\u0159esunout na p\u00e1sku. Toto zdvojen\u00ed z\u00e1lohov\u00e1n\u00ed zvy\u0161uje dobu z\u00e1lohov\u00e1n\u00ed a, co je je\u0161t\u011b d\u016fle\u017eit\u011bj\u0161\u00ed, dobu obnovy. servery<\/a>, OpenShift, Oracle, MariaDB, Cassandra atd. Je dobr\u00e9 z\u00e1lohovat to v\u0161echno spole\u010dn\u00fdm n\u00e1strojem. Instalace samostatn\u00e9ho \u0159e\u0161en\u00ed pouze pro PostgreSQL je \u0161patn\u00fd n\u00e1pad: data se nejprve zkop\u00edruj\u00ed na disk a pot\u00e9 je pot\u0159eba je p\u0159ev\u00e9st na p\u00e1sku. Takov\u00e9 duplicitn\u00ed z\u00e1lohov\u00e1n\u00ed zvy\u0161uje dobu z\u00e1lohov\u00e1n\u00ed a tak\u00e9, co\u017e je mnohem kriti\u010dt\u011bj\u0161\u00ed, \u2013 obnovov\u00e1n\u00ed.<\/p>\n

V enterprise \u0159e\u0161en\u00ed se z\u00e1lohov\u00e1n\u00ed instalace prov\u00e1d\u00ed s ur\u010dit\u00fdm po\u010dtem uzl\u016f vyhrazen\u00e9ho clusteru. Nap\u0159\u00edklad, Commvault dok\u00e1\u017ee pracovat pouze s dvouuzlov\u00fdm clusterem, ve kter\u00e9m jsou Primary a Secondary pevn\u011b p\u0159i\u0159azeny k ur\u010dit\u00fdm uzl\u016fm. A smysl z\u00e1lohovat existuje pouze s Primary, proto\u017ee z\u00e1lohov\u00e1n\u00ed z Secondary m\u00e1 sv\u00e1 omezen\u00ed. Kv\u016fli vlastnostem DBMS se dump na Secondary nevytv\u00e1\u0159\u00ed, a proto zb\u00fdv\u00e1 pouze mo\u017enost souborov\u00e9 z\u00e1lohy.<\/p>\n

Aby se sn\u00ed\u017eila rizika v\u00fdpadku, p\u0159i vytv\u00e1\u0159en\u00ed odoln\u00e9ho syst\u00e9mu vznik\u00e1 \u201e\u017eiv\u00e1\u201c clusterov\u00e1 konfigurace, a Primary m\u016f\u017ee postupn\u011b migrovat mezi r\u016fzn\u00fdmi servery. Nap\u0159\u00edklad software Patroni s\u00e1m spou\u0161t\u00ed Primary na n\u00e1hodn\u011b vybran\u00e9m uzlu clusteru. S\u0158K nem\u00e1 zp\u016fsob, jak to sledovat \u201ez krabice\u201c, a pokud se m\u011bn\u00ed konfigurace, procesy se rozb\u00edjej\u00ed. To znamen\u00e1, \u017ee zaveden\u00ed extern\u00edho \u0159\u00edzen\u00ed br\u00e1n\u00ed S\u0158K efektivn\u011b fungovat, proto\u017ee \u0159\u00eddic\u00ed server jednodu\u0161e nech\u00e1pe, odkud a jak\u00e1 data je t\u0159eba kop\u00edrovat.<\/p>\n

Dal\u0161\u00ed probl\u00e9m \u2013 implementace z\u00e1lohy v Postgres. Ta je mo\u017en\u00e1 p\u0159es dump, a na mal\u00fdch datab\u00e1z\u00edch to funguje. Ale u velk\u00fdch DB dump trv\u00e1 dlouho, vy\u017eaduje mnoho zdroj\u016f a m\u016f\u017ee zp\u016fsobit selh\u00e1n\u00ed instance DB.<\/p>\n

Souborov\u00e1 z\u00e1loha situaci vy\u0159e\u0161\u00ed, ale na velk\u00fdch datab\u00e1z\u00edch prob\u00edh\u00e1 pomalu, proto\u017ee pracuje v jednovl\u00e1knov\u00e9m re\u017eimu. Krom\u011b toho maj\u00ed dodavatel\u00e9 celou \u0159adu dal\u0161\u00edch omezen\u00ed. Nen\u00ed mo\u017en\u00e9 simult\u00e1nn\u011b pou\u017e\u00edvat souborovou a dumpovou z\u00e1lohu, ani nepodporuj\u00ed deduplikaci. Probl\u00e9m\u016f je mnoho, a \u010dasto je jednodu\u0161\u0161\u00ed vybrat m\u00edsto Postgres drahou, ale osv\u011bd\u010denou DB.<\/p>\n

Nen\u00ed kam ustoupit! Za n\u00e1mi je Moskva a v\u00fdvoj\u00e1\u0159i!<\/h2>\n

\nNed\u00e1vno se na\u0161e t\u00fdmu ocitla p\u0159ed nelehk\u00fdm \u00fakolem: v projektu budov\u00e1n\u00ed AIS OSAGo 2.0, kde jsme prov\u00e1d\u011bli IT infrastrukturu, si v\u00fdvoj\u00e1\u0159i pro nov\u00fd syst\u00e9m vybrali PostgreSQL. <\/p>\n

Velk\u00fdm v\u00fdvoj\u00e1\u0159\u016fm softwaru je mnohem snaz\u0161\u00ed pou\u017e\u00edvat \u201em\u00f3dn\u00ed\u201c open-source \u0159e\u0161en\u00ed. V r\u00e1mci Facebooku je dost specialist\u016f, kte\u0159\u00ed podporuj\u00ed pr\u00e1ci tohoto DBMS. A v p\u0159\u00edpad\u011b RSA v\u0161echny \u00fakoly \u201edruh\u00e9ho dne\u201c spo\u010d\u00edvaly na na\u0161ich bedrech. Po\u017eadovalo se od n\u00e1s zajistit odolnost, vytvo\u0159it cluster a samoz\u0159ejm\u011b nastavit z\u00e1lohov\u00e1n\u00ed. Logika postup\u016f byla n\u00e1sleduj\u00edc\u00ed: <\/p>\n