{"id":89038,"date":"2020-07-17T19:42:30","date_gmt":"2020-07-17T17:42:30","guid":{"rendered":"https:\/\/prohoster.info\/blog\/administrirovanie\/logirovanie-v-kubernetes-efk-protiv-plg"},"modified":"2020-07-17T19:42:30","modified_gmt":"2020-07-17T17:42:30","slug":"logirovanie-v-kubernetes-efk-protiv-plg","status":"publish","type":"post","link":"https:\/\/prohoster.info\/sk\/blog\/administrirovanie\/logirovanie-v-kubernetes-efk-protiv-plg","title":{"rendered":"Logovanie v Kubernetes: EFK proti PLG","gt_translate_keys":[{"key":"rendered","format":"text"}]},"content":{"rendered":"

\"Logovanie<\/p>\n

<\/p>\n

Monitorovanie sa stalo ve\u013emi d\u00f4le\u017eitou s\u00fa\u010das\u0165ou rast\u00facich cloudov\u00fdch rie\u0161en\u00ed s rast\u00facou zlo\u017eitost\u00ed distribuovan\u00fdch syst\u00e9mov. Je nevyhnutn\u00e9 na pochopenie ich spr\u00e1vania. Potrebujeme \u0161k\u00e1lovate\u013en\u00e9 n\u00e1stroje, ktor\u00e9 dok\u00e1\u017eu zhroma\u017e\u010fova\u0165 \u00fadaje zo v\u0161etk\u00fdch slu\u017eieb \u2014 a poskytn\u00fa\u0165 odborn\u00edkom jedin\u00e9 rozhranie s anal\u00fdzou v\u00fdkonu, zobrazen\u00edm ch\u00fdb, dostupnos\u0165ou a protokolmi. <\/p>\n

<\/p>\n

Tieto n\u00e1stroje musia by\u0165 efekt\u00edvne a v\u00fdkonn\u00e9. V tomto \u010dl\u00e1nku presk\u00famame dva popul\u00e1rne technologick\u00e9 stacky: EFK (Elasticsearch) a PLG (Loki) a podrobne si objasn\u00edme ich architekt\u00fary a rozdiely.<\/p>\n

<\/a><\/noindex><\/p>\n

Stack EFK<\/h2>\n

<\/p>\n

Mo\u017eno ste u\u017e po\u010duli o ve\u013emi popul\u00e1rnom ELK alebo EFK. Stack sa sklad\u00e1 z viacer\u00fdch samostatn\u00fdch \u010dast\u00ed: Elasticsearch (objektov\u00e9 \u00falo\u017eisko), Logstash alebo FluentD (zhroma\u017e\u010fovanie a agreg\u00e1cia protokolov) a Kibana na vizualiz\u00e1ciu.<\/p>\n

<\/p>\n

Typick\u00e1 sch\u00e9ma pr\u00e1ce vyzer\u00e1 takto:<\/p>\n

<\/p>\n

\"Logovanie<\/p>\n

<\/p>\n

Elasticsearch<\/strong> \u2014 distribuovan\u00e9 objektov\u00e9 \u00falo\u017eisko s vyh\u013ead\u00e1van\u00edm a anal\u00fdzou v re\u00e1lnom \u010dase. Vynikaj\u00face rie\u0161enie pre \u010diasto\u010dne \u0161trukt\u00farovan\u00e9 \u00fadaje, napr\u00edklad protokoly. Inform\u00e1cie sa ukladaj\u00fa vo form\u00e1te JSON, indexuj\u00fa sa v re\u00e1lnom \u010dase a distribuuj\u00fa po uzloch klastra. Pou\u017e\u00edva sa inverzn\u00fd index, ktor\u00fd obsahuje v\u0161etky unik\u00e1tne slov\u00e1 a s nimi s\u00favisiace dokumenty na full-textov\u00e9 vyh\u013ead\u00e1vanie, ktor\u00e9 je zase zalo\u017een\u00e9 na vyh\u013ead\u00e1vacom eng\u00edne Apache Lucene.<\/p>\n

<\/p>\n

FluentD<\/strong> \u2014 to je zberate\u013e \u00fadajov, ktor\u00fd vykon\u00e1va unifik\u00e1ciu \u00fadajov pri ich zbere a spotrebe. Sna\u017e\u00ed sa usporiada\u0165 d\u00e1ta do JSON, ak je to mo\u017en\u00e9. Jeho architekt\u00fara je roz\u0161\u00edrite\u013en\u00e1, existuje viac ako stoviek r\u00f4znych roz\u0161\u00edren\u00ed<\/a><\/noindex>, podporovan\u00fdch komunitou, pre v\u0161etky mo\u017en\u00e9 situ\u00e1cie.<\/p>\n

<\/p>\n

Kibana<\/strong> \u2014 n\u00e1stroj na vizualiz\u00e1ciu \u00fadajov pre Elasticsearch s r\u00f4znymi doplnkov\u00fdmi funkciami, napr\u00edklad anal\u00fdzou \u010dasov\u00fdch radov, grafami, strojov\u00fdm u\u010den\u00edm a \u010fal\u0161\u00edm.<\/p>\n

<\/p>\n

Architekt\u00fara Elasticsearch<\/h3>\n

<\/p>\n

\u00dadaje klastra Elasticsearch s\u00fa ulo\u017een\u00e9 rozpt\u00fdlene po v\u0161etk\u00fdch jeho uzloch. Klastra sa sklad\u00e1 z viacer\u00fdch uzlov na zlep\u0161enie dostupnosti a odolnosti. Ak\u00fdko\u013evek uzol m\u00f4\u017ee vykon\u00e1va\u0165 v\u0161etky \u00falohy klastra, ale v ve\u013ek\u00fdch \u0161k\u00e1lovate\u013en\u00fdch nasadeniach sa uzlom zvy\u010dajne prira\u010fuj\u00fa konkr\u00e9tne \u00falohy.<\/p>\n

<\/p>\n

Typy uzlov klastra:<\/p>\n

<\/p>\n