{"id":36005,"date":"2019-10-31T22:09:06","date_gmt":"2019-10-31T19:09:06","guid":{"rendered":"https:\/\/prohoster.info\/blog\/otkazoustojchivost-v-shd-qsan\/"},"modified":"2019-10-31T22:09:06","modified_gmt":"2019-10-31T19:09:06","slug":"otkazoustojchivost-v-shd-qsan","status":"publish","type":"post","link":"https:\/\/prohoster.info\/sq\/blog\/administrirovanie\/otkazoustojchivost-v-shd-qsan","title":{"rendered":"Fault tolerance in Qsan storage systems","gt_translate_keys":[{"key":"rendered","format":"text"}]},"content":{"rendered":"

Sot dikur n\u00eb infrastruktur\u00ebn IT, me p\u00ebrdorimin e p\u00ebrgjithsh\u00ebm t\u00eb virtualizimit, sistemet e ruajtjes s\u00eb t\u00eb dh\u00ebnave jan\u00eb thelbi q\u00eb ruan t\u00eb gjitha makinat virtuale. D\u00ebshtimi i k\u00ebtij elementi mund t\u00eb ndal\u00eb plot\u00ebsisht funksionin e qendr\u00ebs s\u00eb kujtes\u00ebs. Edhe pse nj\u00eb pjes\u00eb e konsiderueshme e pajisjeve serverike ka d\u00ebshtim t\u00eb q\u00ebndruesh\u00ebm n\u00eb nj\u00eb form\u00eb ose nj\u00eb tjet\u00ebr \"nga-default\", p\u00ebr shkak t\u00eb rolit t\u00eb ve\u00e7ant\u00eb t\u00eb SCHED n\u00eb kuad\u00ebr t\u00eb qendr\u00ebs s\u00eb t\u00eb dh\u00ebnave, k\u00ebrkesat p\u00ebr t\u00eb jan\u00eb m\u00eb t\u00eb larta n\u00eb aspektin e \"vijueshm\u00ebris\u00eb\".<\/p>\n

<\/p>\n

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

Metoda m\u00eb efektive p\u00ebr t\u00eb siguruar d\u00ebshtimin e q\u00ebndruesh\u00ebm n\u00eb IT \u00ebsht\u00eb p\u00ebrdorimi i disa kopjeve t\u00eb pajisjeve dhe SO-s\u00eb (n\u00eb rastin m\u00eb t\u00eb thjesht\u00eb - dyfishimi). Sigurisht, SCHED mund t\u00eb dyfishohet plot\u00ebsisht. Dhe p\u00ebr rikuperimin nga katastrofat, pik\u00ebrisht ky qasje p\u00ebrdoret. Por jo t\u00eb gjith\u00eb kompanit\u00eb mund ta p\u00ebrballojn\u00eb nj\u00eb zgjidhje t\u00eb till\u00eb. Fjal\u00ebn nuk e kam vet\u00ebm p\u00ebr koston e dyfishuar t\u00eb pajisjeve, por edhe p\u00ebr shpenzimet e tjera p\u00ebr organizimin e nj\u00eb zgjidhjeje t\u00eb till\u00eb dhe mb\u00ebshtetjes s\u00eb saj t\u00eb m\u00ebtejshme. <\/p>\n

<\/p>\n

Megjithat\u00eb, mund\u00ebsia e dyfishimit t\u00eb pajisjeve nuk heq nevoj\u00ebn p\u00ebr t\u00eb siguruar d\u00ebshtimin e q\u00ebndruesh\u00ebm n\u00eb nivelin e komponent\u00ebve. N\u00eb ve\u00e7anti, n\u00eb SCHED aplikohet rezervimi p\u00ebr blloqet e energjis\u00eb, modul\u00ebt e ndihm\u00ebs, ruajt\u00ebsit dhe, natyrisht, kontrolluesit. T\u00eb gjitha k\u00ebto kan\u00eb koh\u00eb q\u00eb jan\u00eb b\u00ebr\u00eb rutin\u00eb. \u00cbsht\u00eb e v\u00ebshtir\u00eb t\u00eb gjejm\u00eb SCHED pa p\u00ebrdorimin e nj\u00eb dizajni t\u00eb till\u00eb. Qsan <\/a><\/noindex>k\u00ebtu \u2013 nuk \u00ebsht\u00eb p\u00ebrjashtim. Por d\u00ebshirojm\u00eb t\u00eb flasim n\u00eb k\u00ebt\u00eb artikull p\u00ebr at\u00eb q\u00eb nuk bie n\u00eb sy menj\u00ebher\u00eb, dhe q\u00eb \u00ebsht\u00eb e fokusuar kryesisht n\u00eb rritjen e q\u00ebndrueshm\u00ebris\u00eb s\u00eb sistemit si nj\u00eb e t\u00ebr\u00eb.<\/p>\n

<\/p>\n

Modulet e ndihm\u00ebs<\/h3>\n

<\/p>\n

Shum\u00eb shpesh n\u00eb SCHED <\/a><\/noindex>me kaseta 2U-3U p\u00ebrdoren module t\u00eb kombinuara, q\u00eb bashkojn\u00eb bloket e energjis\u00eb dhe ventilator\u00ebt. Nga nj\u00ebra an\u00eb, kjo \u00ebsht\u00eb e p\u00ebrshtatshme, pasi duhet t\u00eb mir\u00ebmbahen vet\u00ebm nj\u00eb bllok. Nga ana tjet\u00ebr, n\u00ebse sistemi i ndihm\u00ebs d\u00ebshton, mund t\u00eb ndaloj\u00eb bllokun e energjis\u00eb p\u00ebr t\u00eb shmangur mbinxehjen. Dhe duket se do t\u00eb krijohet nj\u00eb situat\u00eb jo aq kritike, por shtimi i dob\u00ebsive n\u00eb SCHED nuk \u00ebsht\u00eb me t\u00eb v\u00ebrtet\u00eb i d\u00ebshiruar.<\/p>\n

<\/p>\n

Ftohja n\u00eb S\u00c7D Qsan \u00ebsht\u00eb organizuar n\u00eb form\u00ebn e moduleve t\u00eb ve\u00e7anta me \"k\u00ebrkes\u00eb t\u00eb nxeht\u00eb\", t\u00eb pavarur nga blloqet e energjis\u00eb. Sakt\u00ebsisht, n\u00eb blloqet e energjis\u00eb ka ventilator\u00eb t\u00eb tyre, t\u00eb destinuar p\u00ebr ajrosjen e vet\u00eb BPs. Moduli i ftohjes p\u00ebrmban dy ventilator\u00eb t\u00eb pavarur, t\u00eb cil\u00ebt sigurojn\u00eb nj\u00ebri-tjetrin. Ka dy module t\u00eb tilla n\u00eb S\u00c7D: nj\u00ebri n\u00eb t\u00eb djatht\u00eb dhe tjetri n\u00eb t\u00eb majt\u00eb \u2013 p\u00ebr ajrosjen efikase t\u00eb t\u00eb gjitha komponenteve. N\u00ebse ndodh nj\u00eb d\u00ebshtim i nj\u00ebrit nga ventilator\u00ebt, t\u00eb gjith\u00eb t\u00eb tjer\u00ebt automatikisht rrisin shpejt\u00ebsin\u00eb e tyre p\u00ebr t\u00eb kompensuar munges\u00ebn e fluksit t\u00eb ajrit q\u00eb \u00ebsht\u00eb krijuar. P\u00ebr k\u00ebt\u00eb arsye, defekti i ventilatorit nuk paraqet rrezik p\u00ebr tepric\u00ebn e gjith\u00eb pajisjes.<\/p>\n

<\/p>\n

Topologjia e lidhjes s\u00eb rafteve t\u00eb zgjerimit<\/h3>\n

<\/p>\n

Skema klasike e lidhjes rafteve t\u00eb zgjerimit<\/a><\/noindex> n\u00eb S\u00c7D n\u00ebnkupton nj\u00eb topologji q\u00eb quhet kaskad\u00eb. N\u00eb k\u00ebt\u00eb rast, kontrollor\u00ebt p\u00ebrkat\u00ebs t\u00eb rafteve dhe S\u00c7D lidhen me nj\u00ebri-tjetrin p\u00ebrmes nj\u00eb kablli SAS t\u00eb vet\u00ebm. Pra, gjithsej rezulton n\u00eb 2 kabuj p\u00ebr nj\u00eb sistem me dy kontrollor\u00eb. N\u00ebse k\u00ebrkohet t\u00eb lidhet e dyta, ajo lidhet nj\u00ebjt\u00eb me rafte e par\u00eb. Dhe k\u00ebshtu me radh\u00eb. Nj\u00eb avantazh i k\u00ebsaj topologjie \u00ebsht\u00eb thjesht\u00ebsia e realizimit n\u00eb pajisje. Nd\u00ebrsa nj\u00eb disavantazh do t\u00eb ishte disa vulnerabilitet ndaj ndarjes papritur t\u00eb zinXhir\u00ebve SAS p\u00ebr shkak t\u00eb d\u00ebshtimit t\u00eb kryq\u00ebzuar t\u00eb kontrollor\u00ebve t\u00eb pand\u00ebr lidhur t\u00eb S\u00c7D dhe rafteve ose p\u00ebr shkak t\u00eb energjis\u00eb s\u00eb humbur t\u00eb nj\u00ebrit nga rafteve t\u00eb zgjerimit n\u00eb mes t\u00eb zinXhir\u00ebve. Si rezultat, do t\u00eb humbasim aksesin n\u00eb nj\u00eb pjes\u00eb t\u00eb ruajt\u00ebsve dhe mund t\u00eb shkat\u00ebrrojm\u00eb grupin RAID, n\u00ebse ai \u00ebsht\u00eb \"i shp\u00ebrndar\u00eb\" n\u00eb disa trupa.<\/p>\n

<\/p>\n

Nga d\u00ebshtimi i kryq\u00ebzuar i kontrollor\u00ebve, Qsan ka nj\u00eb mbrojtje n\u00eb form\u00ebn e nj\u00eb lidhjeje logjike t\u00eb brendshme midis kontrollor\u00ebve p\u00ebrmes backplane t\u00eb S\u00c7D. K\u00ebshtu, kontrollori i S\u00c7D sheh jo vet\u00ebm kontrollorin JBOD, i lidhur drejtp\u00ebrdrejt me t\u00eb, por edhe kontrollorin \"fqinji\" p\u00ebrmes nj\u00eb lidhjeje speciale n\u00eb backplane. Si rezultat, n\u00ebse ndodh kjo situat\u00eb dhe askush fizikisht nuk do t\u00eb t\u00ebrheq\u00eb kabujt SAS midis S\u00c7D dhe rafteve, aksesin n\u00eb t\u00eb gjith\u00eb ruajt\u00ebsit do t\u00eb ruhet.<\/p>\n

<\/p>\n

\"Fault<\/p>\n

P\u00ebr t\u00eb mbrojtur nga nd\u00ebrprerja e zinxhirit SAS, p\u00ebr shembull, p\u00ebr shkak t\u00eb humbjes s\u00eb energjis\u00eb s\u00eb raftit t\u00eb zgjatjes, zakonisht aplikohet nj\u00eb topologji tjet\u00ebr lidh\u00ebse \u2013 kaskada e kund\u00ebrt. N\u00eb k\u00ebt\u00eb rast, SXY lidhet direkt me raftin e par\u00eb dhe t\u00eb fundit n\u00eb zinxhir, duke marr\u00eb qasje n\u00eb disqet si\u00e7 do t\u00eb ishte nga t\u00eb dy an\u00ebt.<\/p>\n

<\/p>\n

\"Fault<\/p>\n

N\u00ebse d\u00ebshiron nj\u00eb mbrojtje m\u00eb t\u00eb fort\u00eb, mund t\u00eb nd\u00ebrtohen konfigurations m\u00eb t\u00eb m\u00ebdha, duke p\u00ebrdorur, p\u00ebr shembull, topologjin\u00eb e pem\u00ebs. Ose mund t\u00eb komplikohet akoma m\u00eb shum\u00eb n\u00ebp\u00ebrmjet kombinimit t\u00eb topologjive t\u00eb p\u00ebrmendura. Kjo \u00ebsht\u00eb e mundur fal\u00eb numrit t\u00eb madh t\u00eb lidhjeve SAS n\u00eb pajisje (2 p\u00ebr \u00e7do kontrollues SXY dhe 5 p\u00ebr \u00e7do kontrollues JBOD) me identifikimin automatik t\u00eb m\u00ebnyrave t\u00eb pun\u00ebs hyrje\/ dalje. E r\u00ebnd\u00ebsishme \u00ebsht\u00eb q\u00eb admini vet\u00eb t\u00eb mos ngat\u00ebrohet. SXY do t\u00eb jet\u00eb n\u00eb gjendje ta konfiguroj\u00eb si\u00e7 duhet.<\/p>\n

<\/p>\n

Fast rebuild<\/h3>\n

<\/p>\n

Prania n\u00eb sistem e disqeve rezerv\u00eb me \u2018z\u00ebvend\u00ebsim t\u00eb nxeht\u00eb\u2019 (hot spare) rrit ndjesh\u00ebm besueshm\u00ebrin\u00eb e ruajtjes s\u00eb informacionit. Megjithat\u00eb, thjesht t\u00eb b\u00ebsh t\u00eb disponuesh\u00ebm disqe t\u00eb till\u00eb nuk do t\u00eb thot\u00eb se ke mbrojtje absolute. Problemi \u00ebsht\u00eb se procesi i rikuperimit (rebild) \u00ebsht\u00eb mjaft i lodhsh\u00ebm dhe shpesh zgjat gjat\u00eb. Kjo v\u00ebshtir\u00ebsi shfaqet p\u00ebr shkak t\u00eb aksesit t\u00eb vazhduesh\u00ebm n\u00eb t\u00eb dh\u00ebnat kryesore. Kjo do t\u00eb thot\u00eb q\u00eb sistemi, p\u00ebrve\u00e7 pun\u00ebs aktuale, duhet gjithashtu t\u00eb kopjoj\u00eb t\u00eb dh\u00ebnat n\u00eb diskun e ri. Dhe gjat\u00ebzgjatja e rebild-it \u00ebsht\u00eb drejtp\u00ebrdrejt e varur nga kapaciteti i disku dhe karakteristikat e tij t\u00eb shpejt\u00ebsis\u00eb. Duke qen\u00eb se sistemi nuk di asgj\u00eb p\u00ebr hap\u00ebsir\u00ebn reale t\u00eb z\u00ebn\u00eb n\u00eb disqe, ai n\u00eb procesin e rebild-it vet\u00ebm kopjon gjith\u00e7ka: bllok pas blloku. <\/p>\n

<\/p>\n

Si rezultat, rikuperimi i nj\u00eb disku modern me kapacitet t\u00eb madh 10+TB n\u00eb ngarkes\u00eb t\u00eb r\u00ebnd\u00eb mbi SXY mund t\u00eb zgjas\u00eb leht\u00ebsisht nj\u00eb jav\u00eb e m\u00eb shum\u00eb. Duhet gjithashtu t\u00eb merret parasysh q\u00eb gjat\u00eb rebild-it, rritet duksh\u00ebm probabiliteti i d\u00ebshtimit t\u00eb disqeve t\u00eb tjera p\u00ebr shkak t\u00eb ngarkes\u00ebs s\u00eb shtuar mbi to. Dhe kjo mund t\u00eb paraqes\u00eb nj\u00eb rrezik serioz n\u00eb rast se p\u00ebrdoret, p\u00ebr shembull, RAID5.<\/p>\n

<\/p>\n

Si zgjidhje p\u00ebr k\u00ebt\u00eb problem, shum\u00eb zhvillues t\u00eb SXY-ve kan\u00eb shqet\u00ebsuar p\u00ebrshpejtimin e procesit t\u00eb rikuperimit. P\u00ebr k\u00ebt\u00eb, mund t\u00eb aplikohen qasje t\u00eb ndryshme, por thelbi mbetet i nj\u00ebjt\u00eb \u2013 kopjimi gjat\u00eb rebild-it i vet\u00ebm bllokave realisht t\u00eb z\u00ebna. Qsan nuk ka mbetur anash nga ky problem. SXY-ja e k\u00ebtij produkti ka nj\u00eb opsion t\u00eb aktivizuar. Fast Rebuild<\/a><\/noindex> sistemi monitoron ndihm\u00ebn e p\u00ebrdorur p\u00ebr regjistrimin e bllok\u00ebve, duke pasur k\u00ebshtu mund\u00ebsin\u00eb q\u00eb n\u00eb rast t\u00eb nj\u00eb d\u00ebshtimi t\u00eb harduerit t\u00eb kopjoj\u00eb vet\u00ebm ata n\u00eb nj\u00eb nj\u00ebsi t\u00eb re.<\/p>\n

<\/p>\n

\"Fault<\/p>\n

Opsioni Fast Rebuild nuk \u00ebsht\u00eb i aktivizuar si parazgjedhje gjat\u00eb krijimit t\u00eb volumieve, pasi p\u00ebrdorimi i tij ka ndikim n\u00eb performanc\u00eb, ve\u00e7an\u00ebrisht gjat\u00eb operacioneve t\u00eb shkruarjes rast\u00ebsore, sepse:<\/p>\n

<\/p>\n

    \n
  1. Duhet t\u00eb monitorohet regjistrimi n\u00eb blloqe;<\/li>\n
  2. Gjat\u00eb ribuild-it nuk ndodh ribarazimi i kontrolleve p\u00ebr hap\u00ebsirat e pap\u00ebrdorura, prandaj p\u00ebr regjistrimin e ri n\u00eb k\u00ebt\u00eb zon\u00eb \u00ebsht\u00eb e nevojshme s\u00eb pari \"t\u00eb inicializohet\" ajo.<\/li>\n<\/ol>\n

    <\/p>\n

    Prandaj, nuk rekomandohet p\u00ebrdorimi i Fast Rebuild p\u00ebr volumet, p\u00ebr shembull, me baza t\u00eb dh\u00ebnash me ngarkes\u00eb t\u00eb lart\u00eb ose n\u00eb sistemi monitorimi, ku volumi gjithsesi do t\u00eb mbushet n\u00eb 100%. Por p\u00ebr server\u00ebt e skedar\u00ebve ose t\u00eb post\u00ebs, ky opsion do t\u00eb jet\u00eb n\u00eb t\u00eb v\u00ebrtet\u00eb shum\u00eb i dobish\u00ebm.<\/p>\n

    <\/p>\n

    N\u00eb p\u00ebrfundim<\/h2>\n

    <\/p>\n

    \u00c7do prodhues i regjistrimeve t\u00eb dh\u00ebnash n\u00ebnkupton se pajisjet e tij jan\u00eb t\u00eb besueshme. Dhe n\u00ebse nuk ka gabime fatale gjat\u00eb zhvillimit t\u00eb pajisjeve dhe p\u00ebrpjekjeve t\u00eb jasht\u00ebzakonshme p\u00ebr kursimin e kostove gjat\u00eb procesit t\u00eb prodhimit dhe testimit, n\u00eb p\u00ebrgjith\u00ebsi, mund t\u00eb bien dakord me shit\u00ebsin. Megjithat\u00eb, duhet t\u00eb kuptohet: <\/p>\n

    <\/p>\n