{"id":95783,"date":"2020-10-03T13:42:02","date_gmt":"2020-10-03T11:42:02","guid":{"rendered":"https:\/\/prohoster.info\/blog\/administrirovanie\/nastrojka-yadra-linux-dlya-glusterfs"},"modified":"2020-10-03T13:42:02","modified_gmt":"2020-10-03T11:42:02","slug":"nastrojka-yadra-linux-dlya-glusterfs","status":"publish","type":"post","link":"https:\/\/prohoster.info\/sq\/blog\/administrirovanie\/nastrojka-yadra-linux-dlya-glusterfs","title":{"rendered":"Konfigurimi i b\u00ebrtham\u00ebs Linux p\u00ebr GlusterFS","gt_translate_keys":[{"key":"rendered","format":"text"}]},"content":{"rendered":"

P\u00ebrkthimi i artikullit \u00ebsht\u00eb p\u00ebrgatitur para fillimit t\u00eb kursit \u00abAdministrator Linux. Profesional\u00bb<\/a><\/noindex>.<\/i><\/b><\/p>\n

\"Konfigurimi<\/p>\n

Her\u00eb pas here shfaqen pyetje mbi rekomandimet e Gluster p\u00ebr konfigurimin e b\u00ebrtham\u00ebs dhe n\u00ebse ka nevoj\u00eb p\u00ebr at\u00eb.<\/p>\n

Kjo nevoj\u00eb ndodh rrall\u00eb. N\u00eb shumic\u00ebn e ngarkesave, b\u00ebrthama funksionon shum\u00eb mir\u00eb. Megjithat\u00eb, ka edhe nj\u00eb an\u00eb tjet\u00ebr. Historikisht, b\u00ebrthama Linux merr shum\u00eb memorie n\u00ebse i jepet mund\u00ebsia, p\u00ebrfshir\u00eb p\u00ebr memorizimin e informacionit si m\u00ebnyra kryesore p\u00ebr t\u00eb p\u00ebrmir\u00ebsuar performanc\u00ebn.<\/a><\/noindex><\/p>\n

N\u00eb shumic\u00ebn e rasteve, kjo punon shum\u00eb mir\u00eb, por me ngarkes\u00eb t\u00eb madhe mund t\u00eb shkaktoj\u00eb probleme. <\/p>\n

Ne kemi nj\u00eb p\u00ebrvoj\u00eb t\u00eb madhe me sisteme q\u00eb konsumojn\u00eb shum\u00eb memorie, si CAD, EDA dhe t\u00eb ngjashme, t\u00eb cilat filluan t\u00eb ngadal\u00ebsohen n\u00ebn ngarkes\u00eb t\u00eb lart\u00eb. Dhe ndonj\u00ebher\u00eb has\u00ebm n\u00eb probleme me Gluster. Pas nj\u00eb v\u00ebzhgimi t\u00eb holl\u00ebsish\u00ebm p\u00ebr disa dit\u00eb mbi p\u00ebrdorimin e memories dhe koh\u00ebn e pritjes s\u00eb disqeve, e kuptuam se kishim mbingarkes\u00eb, pritje t\u00eb madhe t\u00eb i\/o, gabime t\u00eb b\u00ebrtham\u00ebs (kernel oops), ngrirje dhe t\u00eb tjera. <\/p>\n

Ky artikull \u00ebsht\u00eb rezultat i shum\u00eb eksperimenteve p\u00ebr konfigurimin e parametrave, t\u00eb kryera n\u00eb situata t\u00eb ndryshme. Fal\u00eb k\u00ebtyre parametrave, jo vet\u00ebm q\u00eb u p\u00ebrmir\u00ebsua p\u00ebrgjigjshm\u00ebria n\u00eb p\u00ebrgjith\u00ebsi, por gjithashtu u stabilizua ndjesh\u00ebm funksionimi i klasterit.<\/p>\n

Kur b\u00ebhet fjal\u00eb p\u00ebr konfigurimin e memories, gj\u00ebja e par\u00eb q\u00eb duhet par\u00eb \u00ebsht\u00eb n\u00ebn-sistemi i memories virtuale (VM, virtual memory), i cili ka nj\u00eb num\u00ebr t\u00eb madh opsionesh q\u00eb mund t'ju ngat\u00ebrrojn\u00eb.<\/p>\n

vm.swappiness<\/h4>\n

\nParametri vm.swappiness<\/code> kufizon se sa b\u00ebrthama p\u00ebrdor swapimin (swap, p\u00ebrkthim) n\u00eb krahasim me memorien operative. N\u00eb kodin burimor, ai gjithashtu p\u00ebrcaktohet si \u00abtendency to steal mapped memory\u00bb (prirja p\u00ebr t\u00eb vjedhur memorien e mapuar). Nj\u00eb vler\u00eb e lart\u00eb e swappiness do t\u00eb thot\u00eb se b\u00ebrthama do t\u00eb jet\u00eb m\u00eb e prirur p\u00ebr t\u00eb shkarkuar faqet e mapuara. Nj\u00eb vler\u00eb e ul\u00ebt e swappiness do t\u00eb thot\u00eb t\u00eb kund\u00ebrt\u00ebn: b\u00ebrthama do t\u00eb shkarkoj\u00eb m\u00eb pak faqe nga memoria. N\u00eb fjal\u00eb t\u00eb tjera, sa m\u00eb e lart\u00eb t\u00eb jet\u00eb vlera vm.swappiness<\/code>, aq m\u00eb shum\u00eb sistemi do t\u00eb p\u00ebrdor\u00eb swap.<\/p>\n

P\u00ebrdorimi i lart\u00eb i swap-it nuk \u00ebsht\u00eb e d\u00ebshirueshme, pasi ngarkohen dhe shkarkohen blloqe t\u00eb m\u00ebdha t\u00eb dh\u00ebnash n\u00eb memorien operative. Shum\u00eb thon\u00eb se vlera e swapiness duhet t\u00eb jet\u00eb e lart\u00eb, por sipas p\u00ebrvoj\u00ebs time, rritja e performanc\u00ebs arrihet duke vendosur n\u00eb \u00ab0\u00bb.<\/p>\n

M\u00eb shum\u00eb mund t\u00eb lexoni k\u00ebtu \u2014 lwn.net\/Articles\/100978<\/a><\/noindex> <\/p>\n

Por, p\u00ebrs\u00ebri, k\u00ebto cil\u00ebsime duhet t\u00eb aplikohen me kujdes dhe vet\u00ebm pas testimit t\u00eb aplikacionit t\u00eb ve\u00e7ant\u00eb. P\u00ebr aplikacionet me ngarkes\u00eb t\u00eb lart\u00eb t\u00eb transmetimit, ky paramet\u00ebr duhet t\u00eb vendoset n\u00eb \"0\". Duke e ndryshuar n\u00eb \"0\", p\u00ebrgjigjshm\u00ebria e sistemit p\u00ebrmir\u00ebsohet.<\/p>\n

vm.vfs_cache_pressure<\/h4>\n

\nKy paramet\u00ebr kontrollon kujtes\u00ebn e konsumuar nga b\u00ebrthama p\u00ebr t\u00eb ruajtur objekte katalogu dhe indeksin e t\u00eb dh\u00ebnave (dentry dhe inode).<\/p>\n

Me nj\u00eb vler\u00eb parazgjedhje prej 100, b\u00ebrthama do t\u00eb p\u00ebrpiqet t\u00eb liroj\u00eb ndihm\u00ebn e dentry dhe inode n\u00eb m\u00ebnyr\u00eb \"t\u00eb drejt\u00eb\" n\u00eb raport me pagecache dhe swapcache. Reduktimi i vfs_cache_pressure b\u00ebn q\u00eb b\u00ebrthama t\u00eb mbaj\u00eb ndihm\u00ebn e dentry dhe inode. Kur vlera \u00ebsht\u00eb \"0\", b\u00ebrthama nuk do t\u00eb pastroj\u00eb kurr\u00eb ndihm\u00ebn e dentry dhe inode p\u00ebr shkak t\u00eb munges\u00ebs s\u00eb kujtes\u00ebs (memory pressure), dhe kjo mund t\u00eb \u00e7oj\u00eb leht\u00eb n\u00eb nj\u00eb gabim out-of-memory. Rritja e vfs_cache_pressure p\u00ebr m\u00eb shum\u00eb se 100 b\u00ebn q\u00eb b\u00ebrthama t\u00eb jap\u00eb prioritet p\u00ebr shkarkimin e dentry dhe inode.<\/p>\n

Kur p\u00ebrdoret GlusterFS, shum\u00eb p\u00ebrdorues me sasi t\u00eb m\u00ebdha t\u00eb dh\u00ebnash dhe shum\u00eb skedar\u00eb t\u00eb vegj\u00ebl mund t\u00eb p\u00ebrdorin n\u00eb server nj\u00eb sasi t\u00eb konsiderueshme t\u00eb kujtes\u00ebs p\u00ebr shkak t\u00eb ruajtjes s\u00eb inode\/dentry, e cila mund t\u00eb \u00e7oj\u00eb n\u00eb uljen e performanc\u00ebs, pasi b\u00ebrthama duhet t\u00eb p\u00ebrballoj\u00eb strukturat e t\u00eb dh\u00ebnave n\u00eb nj\u00eb sistem me 40 GB kujtes\u00eb. Vendosja e k\u00ebtij parametri mbi 100 ka ndihmuar shum\u00eb p\u00ebrdorues t\u00eb arrijn\u00eb ruajtje m\u00eb t\u00eb drejt\u00eb dhe t\u00eb p\u00ebrmir\u00ebsojn\u00eb p\u00ebrgjigjshm\u00ebrin\u00eb e b\u00ebrthames.<\/p>\n

vm.dirty_background_ratio dhe vm.dirty_ratio<\/h4>\n

\nParametri i par\u00eb (vm.dirty_background_ratio<\/code>) p\u00ebrcakton p\u00ebrqindjen e kujtes\u00ebs me faqe t\u00eb ndyr\u00eb, arritja e t\u00eb cil\u00ebs duhet t\u00eb filloj\u00eb nj\u00eb rikthim n\u00eb sfond t\u00eb faqeve t\u00eb ndyr\u00eb n\u00eb disk. Derisa ky p\u00ebrqindje t\u00eb mos arrihet, faqe nuk do t\u00eb kthehen n\u00eb disk. Dhe kur rikthimi fillon, ai realizohet n\u00eb sfond, pa nd\u00ebrprer\u00eb proceset n\u00eb pun\u00eb. <\/p>\n

Parametri i dyt\u00eb (vm.dirty_ratio<\/code>) p\u00ebrcakton p\u00ebrqindjen e memories q\u00eb mund t\u00eb z\u00ebn\u00eb faqet e ndotura para fillimit t\u00eb shkarkimit t\u00eb detyruesh\u00ebm (forced flash). Kur arrihet ky prag, t\u00eb gjith\u00eb proceset b\u00ebhen sinkronike (blokohen) dhe nuk u lejohet t\u00eb vazhdojn\u00eb pun\u00ebn derisa operacioni i k\u00ebrkuar i hyrjes\/daljes t\u00eb p\u00ebrfundoj\u00eb n\u00eb m\u00ebnyr\u00eb faktike dhe t\u00eb dh\u00ebnat t\u00eb jen\u00eb n\u00eb disk. N\u00eb hyrje\/dalje me ngarkes\u00eb t\u00eb lart\u00eb, kjo shkakton probleme, pasi ndodhet mungesa e cache-it t\u00eb t\u00eb dh\u00ebnave, dhe t\u00eb gjith\u00eb proceset q\u00eb kryejn\u00eb hyrje\/dalje bllokohen duke pritur p\u00ebr hyrje\/dalje. Kjo \u00e7on n\u00eb nj\u00eb num\u00ebr t\u00eb madh procesesh q\u00eb ngecin, ngarkes\u00eb t\u00eb lart\u00eb, funksionim t\u00eb paq\u00ebndruesh\u00ebm t\u00eb sistemit dhe performanc\u00eb t\u00eb dob\u00ebt.<\/p>\n

Reduktimi i vlerave t\u00eb k\u00ebtyre parametrave \u00e7on n\u00eb gjendjen q\u00eb t\u00eb dh\u00ebnat shkarkohen m\u00eb shpesh n\u00eb disk dhe nuk ruajn\u00eb n\u00eb RAM. Kjo mund t\u00eb ndihmoj\u00eb sistemet me shum\u00eb memorie, p\u00ebr t\u00eb cilat \u00ebsht\u00eb normale t\u00eb shkarkojn\u00eb n\u00eb disk cache t\u00eb faqeve me madh\u00ebsi 45\u201390 GB, q\u00eb \u00e7on n\u00eb koh\u00eb t\u00eb m\u00ebdha pritjeje p\u00ebr aplikacionet n\u00eb frontend, duke ulur p\u00ebrgjigjen dhe nd\u00ebrveprimin e p\u00ebrgjithsh\u00ebm.<\/p>\n

\u00ab1\u00bb > \/proc\/sys\/vm\/pagecache<\/h4>\n

\nCache-i i faqeve (page cache) \u00ebsht\u00eb cache ku ruhen t\u00eb dh\u00ebnat e skedareve dhe programeve t\u00eb ekzekutueshme, pra jan\u00eb faqet me p\u00ebrmbajtje faktike t\u00eb skedareve ose pajisjeve t\u00eb bllokut. Kjo cache p\u00ebrdoret p\u00ebr t\u00eb zvog\u00ebluar numrin e leximeve nga disku. Vlera \u00ab1\u00bb do t\u00eb thot\u00eb q\u00eb p\u00ebr cache p\u00ebrdoret 1% e RAM dhe operacionet e leximit nga disku do t\u00eb jen\u00eb m\u00eb t\u00eb shumta se ato nga RAM. Nuk \u00ebsht\u00eb e nevojshme t\u00eb ndryshoni k\u00ebt\u00eb parametrin, por n\u00ebse jeni paranojak n\u00eb lidhje me kontrollin mbi cache-in e faqeve, mund ta p\u00ebrdorni.<\/p>\n

\u00abdeadline\u00bb > \/sys\/block\/sdc\/queue\/scheduler<\/h4>\n

\nPlanifikuesi i hyrje\/daljes (I\/O scheduler) \u00ebsht\u00eb nj\u00eb komponent i b\u00ebrtham\u00ebs Linux, i cili p\u00ebrpunon radh\u00ebt e leximit dhe shkruarjes. N\u00eb teori, p\u00ebr nj\u00eb kontrollues RAID inteligjent, \u00ebsht\u00eb m\u00eb mir\u00eb t\u00eb p\u00ebrdorni \u00abnoop\u00bb, sepse Linux nuk di asgj\u00eb p\u00ebr gjeometrin\u00eb fizike t\u00eb diskut, prandaj \u00ebsht\u00eb m\u00eb efektive t\u00eb lejohet kontrolluesi, i cili e njeh mir\u00eb gjeometrin\u00eb e diskut, t\u00eb p\u00ebrpunoj\u00eb k\u00ebrkes\u00ebn sa m\u00eb shpejt t\u00eb jet\u00eb e mundur. Por duket se \u00abdeadline\u00bb rrit performanc\u00ebn. M\u00eb shum\u00eb rreth planifikuesve mund t\u00eb lexoni n\u00eb dokumentacionin e kodit burimor t\u00eb b\u00ebrtham\u00ebs Linux: linux\/Documentation\/block\/*osched.txt<\/code>. Dhe gjithashtu kam v\u00ebn\u00eb re nj\u00eb rritje t\u00eb kapacitetit t\u00eb leximit gjat\u00eb operacioneve t\u00eb p\u00ebrziera (shum\u00eb operacione shkrimi).<\/p>\n

\u00ab256\u00bb > \/sys\/block\/sdc\/queue\/nr_requests<\/h4>\n

\nNumri i k\u00ebrkesave t\u00eb hyrje-dalih n\u00eb buffer para se ato t'i kalohen planifikuesit. Madh\u00ebsia e radh\u00ebs s\u00eb brendshme t\u00eb disa kontroler\u00ebve (queue_depth) \u00ebsht\u00eb m\u00eb e madhe se nr_requests e planifikuesit t\u00eb hyrje-dalih, k\u00ebshtu q\u00eb planifikuesi ka pak mund\u00ebsi p\u00ebr t\u00eb priorizuar dhe realizuar bashkimin e k\u00ebrkesave n\u00eb m\u00ebnyr\u00eb t\u00eb sakt\u00eb. P\u00ebr planifikuesit deadline dhe CFQ, \u00ebsht\u00eb m\u00eb mir\u00eb q\u00eb nr_requests t\u00eb jet\u00eb dy her\u00eb m\u00eb i madh se radh\u00ebs s\u00eb brendshme t\u00eb kontroler\u00ebve. Bashkimi dhe riorganizimi i k\u00ebrkesave ndihmojn\u00eb planifikuesin t\u00eb jet\u00eb m\u00eb reagues gjat\u00eb ngarkesave t\u00eb m\u00ebdha.<\/p>\n

echo \u00ab16\u00bb > \/proc\/sys\/vm\/page-cluster<\/h4>\n

\nParametri page-cluster menaxhon numrin e faqeve q\u00eb shkruhen n\u00eb swap nj\u00eb her\u00eb. N\u00eb shembullin e m\u00ebsip\u00ebrm, vlera vendoset n\u00eb \u00ab16\u00bb n\u00eb p\u00ebrputhje me madh\u00ebsin\u00eb e striptit (stripe size) t\u00eb RAID prej 64 KB. Kjo nuk ka ndonj\u00eb kuptim kur swappiness = 0, por n\u00ebse e ke vendosur swappiness n\u00eb 10 ose 20, p\u00ebrdorimi i k\u00ebsaj vlere do t'ju ndihmoj\u00eb kur madh\u00ebsia e striptit RAID \u00ebsht\u00eb 64 KB.<\/p>\n

blockdev \u2014setra 4096 \/dev\/<<\/code>emri_ijon><\/code> (-sdb, hdc ose dev_mapper)<\/h4>\n

\nCil\u00ebsimet e pajisjeve bllok po default p\u00ebr shum\u00eb kontroler\u00eb RAID shpesh \u00e7ojn\u00eb n\u00eb performanc\u00eb t\u00eb tmerrshme. Shtimi i opsionit t\u00eb sip\u00ebrp\u00ebrmendur, konfiguroni leximin parashikues p\u00ebr 4096 * 512-byte sektor\u00eb. T\u00eb pakt\u00ebn p\u00ebr operacionet e rrjedh\u00ebs, rrit efikasitetin duke mbushur \u043a\u0435\u0448in e brendsh\u00ebm t\u00eb disks me leximin parashikues gjat\u00ebsin\u00eb e koh\u00ebs q\u00eb b\u00ebrthama p\u00ebrgatit hyrje-dalih. N\u00eb \u043a\u0435\u0448 mund t\u00eb vendosen t\u00eb dh\u00ebnat q\u00eb do t\u00eb k\u00ebrkohen gjat\u00eb leximit t\u00eb ardhsh\u00ebm. Leximi parashikues shum\u00eb i madh mund t\u00eb shkat\u00ebrroj\u00eb hyrje-dalihin rast\u00ebsor p\u00ebr skedar\u00ebt e m\u00ebdhenj, n\u00ebse ai p\u00ebrdor potencialisht koh\u00ebn e dobishme t\u00eb disks ose ngarkon t\u00eb dh\u00ebna p\u00ebrtej \u043a\u0435\u0448it.<\/p>\n

M\u00eb posht\u00eb jan\u00eb disa rekomandime t\u00eb tjera n\u00eb nivelin e sistemit t\u00eb fileve. Por ato ende nuk jan\u00eb testuar. Sigurohuni q\u00eb sistemi juaj i skedar\u00ebve e di madh\u00ebsin\u00eb e striptit dhe numrin e disqeve n\u00eb grup. P\u00ebr shembull, q\u00eb ky \u00ebsht\u00eb nj\u00eb grup raid5 me madh\u00ebsin\u00eb e striptit 64K nga gjasht\u00eb disqe (me t\u00eb v\u00ebrtet\u00eb nga pes\u00eb, sepse nj\u00eb disk p\u00ebrdoret p\u00ebr paritet). K\u00ebto rekomandime jan\u00eb t\u00eb bazuara n\u00eb supozime teorike dhe jan\u00eb mbledhur nga blogje\/artikuj t\u00eb ekspert\u00ebve t\u00eb RAID.<\/p>\n

-> ext4 fs, 5 disqe, 64K stripe, nj\u00ebsit\u00eb n\u00eb 4K blloqe\nmkfs -text4 -E stride=$((64\/4))\n-> xfs, 5 disqe, 64K stripe, nj\u00ebsit\u00eb n\u00eb sektor\u00eb 512-byte\nmkfs -txfs -d sunit=$((64*2)) -d swidth=$((5*64*2))<\/code><\/pre>\n
\nP\u00ebr skedar\u00ebt e m\u00ebdhenj, mund t\u00eb shqyrtoni mund\u00ebsin\u00eb p\u00ebr t\u00eb rritur p\u00ebrmasat e rreshtave t\u00eb p\u00ebrmendura m\u00eb sip\u00ebr.<\/p>\n
K\u00cbSHILL\u00cb!<\/b> Gjith\u00e7ka q\u00eb \u00ebsht\u00eb p\u00ebrshkruar m\u00eb sip\u00ebr \u00ebsht\u00eb jasht\u00ebzakonisht subjektive p\u00ebr disa lloje aplikacionesh. Ky artikull nuk garanton p\u00ebrmir\u00ebsime pa testimin paraprak t\u00eb aplikacioneve p\u00ebrkat\u00ebse nga p\u00ebrdoruesi. Ai duhet t\u00eb p\u00ebrdoret vet\u00ebm n\u00eb rast se p\u00ebrmir\u00ebson p\u00ebrgjigjen e p\u00ebrgjithshme t\u00eb sistemit ose n\u00ebse zgjidh problemet aktuale.<\/p>\n
Materiale Shtes\u00eb:<\/h4>\n
<\/p>\n