Kuv xav kom koj nyeem cov ntawv sau tseg ntawm tsab ntawv ceeb toom txij thaum pib ntawm 2019 los ntawm Andrey Borodin "Cov ntaub ntawv thaub qab nrog WAL-G. Puas muaj dab tsi hauv 2019?"
Nyob zoo sawv daws! Kuv lub npe yog Andrey Borodin. Kuv yog tus tsim tawm ntawm Yandex. Kuv tau txaus siab rau PostgreSQL txij li xyoo 2016, tom qab kuv tham nrog cov neeg tsim khoom thiab lawv tau hais tias txhua yam yog yooj yim - koj nqa cov cai thiab tsim nws, thiab txhua yam yuav ua haujlwm. Thiab txij thaum ntawd los kuv tsis tuaj yeem nres - Kuv sau txhua yam sib txawv.
Ib yam uas kuv tab tom ua haujlwm yog lub kaw lus thaub qab. . Feem ntau, ntawm Yandex peb tau ua haujlwm ntawm cov txheej txheem thaub qab hauv PostgreSQL tau ntev heev. Thiab koj tuaj yeem pom hauv Is Taws Nem ntau ntawm rau daim ntawv qhia txog yuav ua li cas peb ua cov txheej txheem thaub qab. Thiab txhua xyoo lawv hloov zuj zus me ntsis, txhim kho me ntsis, thiab ua kom ntseeg tau ntau dua.
Tab sis niaj hnub no tsab ntawv ceeb toom tsis yog tsuas yog hais txog qhov peb tau ua, tab sis kuj hais txog qhov yooj yim npaum li cas thiab yog dab tsi. Muaj pes tsawg tus ntawm koj twb tau saib kuv cov lus ceeb toom txog WAL-G? Nws yog qhov zoo uas ob peb tus neeg tsis tau saib, vim kuv yuav pib nrog qhov yooj yim tshaj plaws.
Yog tias mam li nco dheev koj muaj PostgreSQL pawg, thiab kuv xav tias txhua tus muaj ob peb nrog lawv, thiab mam li nco dheev tsis muaj cov txheej txheem thaub qab, ces koj yuav tsum tau txais ib qho S3 cia lossis Google Cloud tshaj cia.
Piv txwv li, koj tuaj yeem tuaj rau peb lub rooj muag khoom thiab coj tus lej tshaj tawm rau Yandex Object Storage, uas yog S3 tshaj.
Ces tsim ib lub thoob. Nws tsuas yog lub thawv rau cov ntaub ntawv.
Tsim ib tus neeg siv kev pabcuam.
Tsim tus yuam sij nkag rau cov neeg siv kev pabcuam: aws-s3-key.
Download tau qhov tseeb tso tawm ruaj khov ntawm WAL-G.
Peb cov kev tso tawm ua ntej sib txawv li cas ntawm kev tshaj tawm? Kuv feem ntau hais kom tso ntxov. Thiab yog tias tsis muaj kab laum hauv version rau lub sijhawm txaus, piv txwv li, ib hlis, ces kuv tso nws. Nov yog qhov kev tso tawm no txij lub Kaum Ib Hlis. Thiab qhov no txhais tau hais tias txhua lub hlis peb pom qee yam kab mob, feem ntau hauv kev ua haujlwm tsis tseem ceeb, tab sis peb tseem tsis tau tso tawm. Cov yav dhau los version tsuas yog Kaum Ib Hlis. Tsis muaj kab laum paub rau peb hauv nws, piv txwv li kab mob tau ntxiv raws li qhov project tau ua tiav.
Thaum koj tau rub tawm WAL-G, koj tuaj yeem khiav ib qho yooj yim "cov npe thaub qab" hais kom ua, dhau ntawm ib puag ncig hloov pauv. Thiab nws yuav txuas rau Object Storage thiab qhia koj tias koj muaj dab tsi backups. Thaum xub thawj, tau kawg, koj yuav tsum tsis txhob muaj thaub qab. Lub ntsiab lus ntawm daim slide no yog qhia tias txhua yam yooj yim heev. Qhov no yog console hais kom ua uas lees txais ib puag ncig hloov pauv thiab ua tiav cov lus txib.
Tom qab no, koj tuaj yeem ua koj thawj thaub qab. Hais " thaub qab-push" hauv WAL-G thiab qhia hauv WAL-G qhov chaw pgdata ntawm koj pawg. Thiab feem ntau yuav, PostgreSQL yuav qhia rau koj, yog tias koj tsis tau muaj lub kaw lus thaub qab, uas koj yuav tsum tau ua kom "archive-hom".
Qhov no txhais tau tias koj yuav tsum tau mus rau qhov chaw thiab qhib "archive_mode = on" thiab ntxiv "archive_command", uas kuj yog ib qho lus txib hauv WAL-G. Tab sis rau qee qhov laj thawj tib neeg feem ntau siv cov ntawv sau rau ntawm lub ncauj lus no thiab qhwv nws ncig WAL-G. Thov tsis txhob ua qhov no. Siv cov haujlwm muaj nyob hauv WAL-G. Yog tias koj tsis muaj dab tsi, sau rau . WAL-G xav tias nws yog tib txoj haujlwm uas khiav hauv archive_command.
Peb siv WAL-G feem ntau los tsim kom muaj pawg muaj nyob hauv Yandex Database tswj.
Thiab nws yog feem ntau siv nyob rau hauv ib tug topology ntawm ib tug Master thiab ntau replications. Nyob rau tib lub sijhawm, nws ua ib daim ntawv luam theej hauv Yandex Object Storage.
Cov xwm txheej feem ntau yog tsim cov ntawv theej ntawm pawg siv Point hauv lub sijhawm rov qab los. Tab sis nyob rau hauv cov ntaub ntawv no, kev ua tau zoo ntawm lub backup system tsis yog li ntawd tseem ceeb rau peb. Peb tsuas yog xav tau upload ib pawg tshiab los ntawm kev thaub qab.
Feem ntau, peb xav tau kev ua haujlwm thaub qab thaum ntxiv qhov tshiab. Vim li cas thiaj tseem ceeb? Feem ntau cov tib neeg ntxiv cov node tshiab rau ib pawg vim hais tias pawg uas twb muaj lawm tsis tuaj yeem tuav cov ntawv nyeem. Lawv yuav tsum tau ntxiv ib tug tshiab replica. Yog tias peb ntxiv cov load los ntawm pg_basebackup rau tus tswv, ces tus Xib Hwb yuav tawg. Yog li ntawd, nws yog ib qho tseem ceeb heev rau peb uas peb yuav sai sai upload ib tug tshiab node los ntawm lub archive, tsim tsawg kawg nkaus load ntawm tus tswv.
Thiab lwm qhov xwm txheej zoo sib xws. Qhov no yog qhov yuav tsum tau rov pib dua tus qub Master tom qab hloov Cluster Master los ntawm Data Center uas txuas tau ploj lawm.
- Yog li ntawd, thaum tsim cov kev cai rau cov ntawv luam, peb pom tau tias pg_basebackup tsis haum rau peb thaum ua haujlwm hauv huab.
- Peb xav kom muaj peev xwm compress peb cov ntaub ntawv. Tab sis yuav luag txhua qhov kev thaub qab uas tsis yog qhov tuaj hauv lub thawv yuav muab cov ntaub ntawv compression.
- Peb xav ua kom sib luag txhua yam vim tias tus neeg siv hauv huab yuav ntau tus processor cores. Tab sis yog hais tias peb tsis muaj parallelism nyob rau hauv ib co kev khiav hauj lwm, ces ib tug loj tus naj npawb ntawm cores yuav siv tsis tau.
- Peb xav tau encryption vim feem ntau cov ntaub ntawv tsis yog peb li thiab tsis tuaj yeem khaws cia hauv cov ntawv ntshiab. Los ntawm txoj kev, peb txoj kev koom tes rau WAL-G pib nrog kev nkag mus. Peb ua tiav qhov encryption hauv WAL-G, tom qab ntawd peb tau nug: "Tej zaum ib tus ntawm peb yuav tsim qhov project?" Thiab txij thaum ntawd los kuv tau ua haujlwm nrog WAL-G rau ntau tshaj ib xyoos.
- Peb kuj xav tau kev txhawb nqa cov peev txheej, vim tias dhau sijhawm siv huab, peb pom tias qee zaum tib neeg muaj khoom noj khoom haus tseem ceeb thaum hmo ntuj thiab cov khoom thauj no tsis tuaj yeem cuam tshuam nrog. Tias yog vim li cas peb ntxiv cov peev txheej throttling.
- Raws li kev teev npe thiab kev tswj hwm.
- Thiab pov thawj.
Peb saib ntau yam cuab yeej sib txawv. Hmoov zoo, peb muaj kev xaiv loj hauv PostgreSQL. Thiab txhua qhov chaw peb tau ploj lawm ib yam dab tsi, qee qhov haujlwm me me, qee qhov me me.
Thiab tau tshuaj xyuas cov txheej txheem uas twb muaj lawm, peb tuaj txog qhov xaus tias peb yuav tsim WAL-G. Nws yog ib qhov project tshiab thaum ntawd. Nws yog ib qho yooj yim heev los cuam tshuam txoj kev loj hlob ntawm huab infrastructure ntawm lub backup system.
Lub tswv yim tseem ceeb uas peb ua raws li WAL-G yuav tsum yooj yim li balalaika.
WAL-G muaj 4 cov lus txib. Qhov no:
WAL-PUSH - archive tus ncej.
WAL-FETCH - tau txais lub taub hau.
BACKUP-PUSH - ua ib qho thaub qab.
BACKUP-FETCH - tau ib qho thaub qab los ntawm cov txheej txheem thaub qab.
Qhov tseeb, WAL-G kuj muaj kev tswj hwm cov thaub qab no, piv txwv li sau npe thiab tshem tawm cov ntaub ntawv thiab cov ntaub ntawv thaub qab hauv keeb kwm uas tsis xav tau ntxiv lawm tam sim no.
Ib qho ntawm cov haujlwm tseem ceeb rau peb yog kev ua haujlwm ntawm kev tsim cov ntawv luam delta.
Cov ntawv luam ntawm Delta txhais tau hais tias peb tsis tsim cov ntaub ntawv thaub qab ntawm tag nrho pawg, tab sis tsuas yog cov nplooj ntawv hloov pauv ntawm cov ntaub ntawv hloov pauv hauv pawg. Nws yuav zoo li tias qhov ua haujlwm no zoo ib yam li lub peev xwm rov qab siv WAL. Tab sis peb tuaj yeem dov WAL ib leeg-xov delta thaub qab nyob rau tib lub sijhawm. Raws li, thaum peb muaj ib qho yooj yim thaub qab ua rau hnub Saturday, delta backups txhua hnub, thiab hnub Thursday peb poob, ces peb yuav tsum yob li 4 delta backups thiab 10 teev ntawm WAL. Nws yuav siv sij hawm tib lub sij hawm vim hais tias cov delta backups yob nyob rau hauv parallel.
LSN-raws li deltas - qhov no txhais tau hais tias thaum tsim ib qho thaub qab, peb yuav tsum tau muab txhua nplooj ntawv thiab xyuas nws LSN nrog LSN ntawm cov thaub qab yav dhau los txhawm rau nkag siab tias nws tau hloov pauv. Txhua nplooj ntawv uas tuaj yeem muaj cov ntaub ntawv hloov pauv yuav tsum muaj nyob hauv delta thaub qab.
Raws li kuv tau hais, ntau qhov kev saib xyuas tau them rau kev sib luag.
Tab sis qhov archive API hauv PostgreSQL zoo ib yam. PostgreSQL archives ib daim ntawv WAL thiab thaum rov kho nws thov ib daim ntawv WAL. Tab sis thaum cov ntaub ntawv thov ib daim ntawv WAL siv "WAL-FETCH" hais kom ua, peb hu rau "WAL-PREFETCH" hais kom ua, uas npaj cov ntaub ntawv 8 tom ntej mus nqa cov ntaub ntawv los ntawm cov khoom khaws cia hauv tib lub sijhawm.
Thiab thaum lub database hais kom peb khaws ib cov ntaub ntawv, peb saib archive_status thiab saib seb puas muaj lwm cov ntaub ntawv WAL. Thiab peb kuj tseem sim rub tawm WAL ua ke. Qhov no muab qhov kev ua tau zoo tseem ceeb thiab txo qhov kev ncua deb ntawm tus lej ntawm WALs uas tsis tau khaws cia. Ntau tus neeg tsim khoom lag luam thaub qab ntseeg tias qhov no yog qhov kev pheej hmoo zoo li no vim tias peb vam khom peb cov kev paub ntawm cov cai hauv cov cai uas tsis yog PostgreSQL API. PostgreSQL tsis tau lees tias muaj cov ntaub ntawv archive_status rau peb thiab tsis lees paub cov ntsiab lus, muaj cov cim qhia txog kev npaj rau WAL cov ntaub ntawv nyob ntawd. Txawm li cas los xij, peb tab tom kawm txog qhov chaws, peb pom tias qhov no yog li ntawd thiab peb tab tom sim siv nws. Thiab peb tswj cov kev taw qhia uas PostgreSQL tab tom txhim kho; yog tias tam sim no lub tshuab no tawg, peb yuav tsum tsis txhob siv nws.
Nyob rau hauv nws daim ntawv ntshiab, LSN-raws li WAL delta yuav tsum tau nyeem cov ntaub ntawv pawg uas nws hom-lub sij hawm nyob rau hauv cov ntaub ntawv system tau hloov txij li thaum yav dhau los thaub qab. Peb nyob nrog qhov no tau ntev, yuav luag ib xyoos. Thiab thaum kawg peb tuaj txog qhov xaus tias peb muaj WAL deltas.
Qhov no txhais tau hais tias txhua zaus peb archive WAL ntawm tus Tswv, peb tsis tsuas yog compress nws, encrypt nws thiab xa mus rau lub network, tab sis peb kuj nyeem nws tib lub sijhawm. Peb txheeb xyuas thiab nyeem cov ntaub ntawv hauv nws. Peb nkag siab cov blocks twg tau hloov pauv thiab sau cov ntaub ntawv delta.
Cov ntaub ntawv delta piav qhia txog qee yam ntawm WAL cov ntaub ntawv, piav qhia txog cov blocks twg tau hloov pauv hauv qhov ntau ntawm WAL. Thiab tom qab ntawd cov ntaub ntawv delta no kuj tau khaws cia.
Ntawm no peb tau ntsib nrog qhov tseeb tias peb sib piv txhua yam sai sai, tab sis peb tsis tuaj yeem nyeem cov keeb kwm sib txuas ua ke, vim tias nyob rau qee ntu peb yuav ntsib qhov kawg ntawm WAL cov ntaub ntawv dhau los, uas peb tsis muaj dab tsi txuas nrog, vim Kev nyeem ntawv sib piv coj mus rau qhov peb xub txheeb xyuas yav tom ntej, uas tseem tsis tau muaj yav dhau los.
Raws li qhov tshwm sim, peb yuav tsum muab cov ntawv tsis nkag siab rau hauv _delta_partial cov ntaub ntawv. Yog li ntawd, thaum peb rov qab mus rau yav dhau los, peb yuav muab cov ntaub ntawv ntawm WAL rau hauv ib qho, tom qab ntawd peb yuav txheeb xyuas nws thiab nkag siab txog qhov hloov pauv hauv nws.
Yog hais tias nyob rau hauv keeb kwm ntawm peb ncej parsing muaj tsawg kawg yog ib tug taw tes uas peb tsis to taub dab tsi tshwm sim, ces, raws li, nyob rau hauv lub tom ntej backup peb yuav raug yuam kom nyeem tag nrho pawg dua, ib yam li peb tau ua nrog ib txwm LSN. - raws delta.
Yog li ntawd, tag nrho peb cov kev txom nyem tau coj mus rau qhov tseeb hais tias peb qhib-qhov chaw ntawm WAL-G parsing library. Raws li kuv paub mas tseem tsis tau muaj leej twg siv, tab sis yog leej twg xav tau, sau thiab siv, nws yog nyob rau hauv pej xeem sau. (Txoj kev txuas tshiab )
Yog li ntawd, tag nrho cov ntaub ntawv ntws zoo li nyuaj heev. Peb tus Xib Hwb archives cov ncej thiab archives delta cov ntaub ntawv. Thiab cov replica uas ua rau backup daim ntawv yuav tsum tau txais cov ntaub ntawv delta thaum lub sij hawm uas tau dhau los ntawm backups. Nyob rau hauv cov ntaub ntawv no, ib feem ntawm cov keeb kwm yuav tsum tau ntxiv nyob rau hauv bulk thiab parsed, vim hais tias tsis tag nrho cov keeb kwm fits rau hauv loj segments. Thiab tsuas yog tom qab qhov no tus replica tuaj yeem khaws tag nrho delta thaub qab.
Ntawm cov duab txhua yam zoo li yooj yim dua. Nov yog qhov rub tawm los ntawm ib qho ntawm peb pawg tiag tiag. Peb muaj LSN-raws li, ua nyob rau hauv ib hnub. Thiab peb pom tias LSN-raws li delta backup tau khiav ntawm peb thaum sawv ntxov mus rau tsib thaum sawv ntxov. Qhov no yog lub load nyob rau hauv tus naj npawb ntawm processor cores. WAL-delta tau coj peb mus txog 20 feeb ntawm no, uas yog, nws tau dhau los ua nrawm dua, tab sis tib lub sijhawm muaj kev sib pauv hnyav dua hauv lub network.
Txij li thaum peb muaj cov ntaub ntawv hais txog cov blocks twg tau hloov pauv thiab lub sijhawm twg hauv keeb kwm ntawm cov ntaub ntawv, peb tau mus ntxiv thiab txiav txim siab los koom ua ke ua haujlwm - PostgreSQL txuas ntxiv hu ua "pg_prefaulter"
Qhov no txhais tau hais tias thaum lub hauv paus stand-by ua tiav cov lus txib rov qab, nws qhia WAL-G kom nqa cov ntaub ntawv WAL tom ntej. Peb nkag siab txog qhov twg cov ntaub ntawv thaiv cov txheej txheem rov qab WAL yuav nkag mus rau yav tom ntej thiab pib ua haujlwm nyeem ntawm cov blocks no. Qhov no tau ua tiav txhawm rau txhawm rau nce kev ua haujlwm ntawm SSD cov tswj hwm. Vim tias WAL yob yuav ncav cuag nplooj ntawv uas yuav tsum tau hloov pauv. Nplooj ntawv no yog nyob rau hauv disk thiab tsis nyob rau hauv nplooj ntawv cache. Thiab nws yuav tos synchronously rau nplooj ntawv no tuaj txog. Tab sis nyob ze yog WAL-G, uas paub tias tom ntej ob peb puas megabytes ntawm WAL peb yuav xav tau qee nplooj ntawv thiab tib lub sijhawm pib sov lawv. Initiates ntau disk nkag mus kom lawv raug tua nyob rau hauv parallel. Qhov no ua haujlwm zoo ntawm SSD drives, tab sis, hmoov tsis, nws yog kiag li tsis muaj feem xyuam rau lub hard drive, vim hais tias peb tsuas cuam tshuam nrog nws nrog peb prompts.
Qhov no yog dab tsi nyob rau hauv lub code tam sim no.
Muaj cov yam ntxwv uas peb xav ntxiv.
Daim duab no qhia tau hais tias WAL-delta siv sijhawm luv luv. Thiab qhov no yog nyeem cov kev hloov pauv uas tshwm sim hauv cov ntaub ntawv thaum nruab hnub. Peb tuaj yeem ua WAL-delta tsis yog thaum hmo ntuj, vim nws tsis yog qhov tseem ceeb ntawm kev thauj khoom. Peb tuaj yeem nyeem WAL-delta txhua feeb vim nws pheej yig. Hauv ib feeb peb tuaj yeem tshawb xyuas txhua qhov kev hloov pauv uas tau tshwm sim rau pawg. Thiab qhov no tuaj yeem hu ua "instant WAL-delta".
Lub ntsiab lus yog tias thaum peb rov kho cov pawg, peb txo tus naj npawb ntawm cov dab neeg uas peb yuav tsum tau dov mus ua ntu zus. Ntawd yog, tus nqi ntawm WAL uas PostgreSQL yob yuav tsum raug txo, vim nws yuav siv sij hawm tseem ceeb.
Tab sis tsis yog tag nrho. Yog tias peb paub tias qee qhov thaiv yuav raug hloov mus rau qhov sib npaug ntawm qhov sib npaug, peb tsis tuaj yeem hloov pauv yav dhau los. Ntawd yog, tam sim no peb muaj cov ntaub ntawv-ntawm-cov ntaub ntawv optimization ntawm WAL-delta forwarding. Qhov no txhais tau hais tias, piv txwv li, hnub Tuesday lub rooj raug tshem tawm tag lossis qee cov ntaub ntawv raug tshem tawm tag nrho ntawm lub rooj, tom qab ntawd thaum delta dov rau hnub Monday thiab Saturday lub pg_basebackup rov qab los, peb yuav tsis tsim cov ntaub ntawv no.
Peb xav txuas ntxiv cov thev naus laus zis no mus rau qib nplooj ntawv. Ntawd yog, yog tias qee qhov ntawm cov ntaub ntawv hloov pauv rau hnub Monday, tab sis yuav raug sau dua rau hnub Wednesday, tom qab ntawd thaum rov qab mus rau qhov taw tes rau hnub Thursday, peb tsis tas yuav sau thawj ob peb nplooj ntawv rau disk.
Tab sis qhov no tseem yog ib lub tswv yim uas tab tom tham txog nyob rau hauv peb, tab sis nws tseem tsis tau mus txog qhov chaws.
Peb xav ua ib qho ntxiv hauv WAL-G. Peb xav ua kom nws nthuav dav vim tias peb yuav tsum txhawb nqa cov ntaub ntawv sib txawv thiab xav kom muaj peev xwm ua kom muaj kev tswj hwm thaub qab tib yam. Tab sis qhov teeb meem yog tias MySQL APIs txawv heev. Hauv MySQL, PITR tsis yog nyob ntawm lub cev WAL, tab sis ntawm binlog. Thiab peb tsis muaj archiving system nyob rau hauv MySQL uas yuav qhia ib co sab nraud system tias qhov binlog no tiav lawm thiab yuav tsum tau archived. Peb yuav tsum tau sawv ntawm qhov chaw hauv cron nrog cov ntaub ntawv thiab xyuas seb puas muaj ib yam dab tsi npaj tau?
Thiab tib txoj kev, thaum lub sij hawm MySQL restore, tsis muaj cov lus txib rov qab uas tuaj yeem qhia rau lub kaw lus uas kuv xav tau cov ntaub ntawv zoo li no. Ua ntej koj pib rov tsim koj pawg, koj yuav tsum paub cov ntaub ntawv koj xav tau. Koj tus kheej yuav tsum twv seb cov ntaub ntawv twg koj yuav xav tau. Tab sis cov teeb meem no yuav muaj peev xwm zam tau qee yam. (Kev qhia meej: MySQL twb tau txais kev txhawb nqa)
Hauv tsab ntawv ceeb toom, kuv kuj xav tham txog cov xwm txheej thaum WAL-G tsis haum rau koj.
Yog tias koj tsis muaj qhov hloov pauv synchronous, WAL-G tsis lav tias qhov kawg ntu yuav raug khaws cia. Thiab yog tias archiving lags qab ob peb ntu kawg ntawm keeb kwm, qhov ntawd yog qhov txaus ntshai. Yog tias tsis muaj qhov hloov pauv synchronous, kuv yuav tsis pom zoo siv WAL-G. Tseem, nws tsuas yog tsim los rau kev teeb tsa huab, uas txhais tau hais tias muaj kev daws teeb meem siab nrog lub synchronous replica, uas yog lub luag haujlwm rau kev nyab xeeb ntawm qhov kawg bytes tau cog lus.
Kuv feem ntau pom cov neeg sim khiav ob qho tib si WAL-G thiab WAL-E tib lub sijhawm. Peb txhawb kev rov qab sib raug zoo hauv kev nkag siab tias WAL-G tuaj yeem rov qab tau cov ntaub ntawv los ntawm WAL-E thiab tuaj yeem rov qab tau cov ntaub ntawv thaub qab hauv WAL-E. Tab sis txij li ob qho tib si ntawm cov tshuab no siv tib lub sijhawm wal-push, lawv pib nyiag cov ntaub ntawv los ntawm ib leeg. Yog tias peb kho nws hauv WAL-G, nws tseem yuav nyob hauv WAL-E. Hauv WAL-E, nws saib ntawm archive-status, pom cov ntaub ntawv tiav thiab khaws cia lawv, thaum lwm lub tshuab tsuas yuav tsis paub tias WAL cov ntaub ntawv no muaj, vim PostgreSQL yuav tsis sim archive nws zaum ob.
Peb yuav kho dab tsi ntawm no ntawm WAL-G sab? Peb yuav tsis qhia PostgreSQL tias cov ntaub ntawv no tau hloov mus ua ke, thiab thaum PostgreSQL nug peb kom khaws nws, peb yuav twb paub tias cov ntaub ntawv zoo li no nrog lub sijhawm-lub sijhawm thiab nrog md5 no twb tau muab khaws cia thiab peb yuav tsuas hais PostgreSQL - OK, txhua yam yog npaj txhij yam tsis tseem ceeb ua dab tsi.
Tab sis qhov teeb meem no tsis zoo li yuav raug kho ntawm WAL-E sab, yog li tam sim no nws tsis tuaj yeem tsim cov lus txib archive uas yuav khaws cov ntaub ntawv hauv WAL-G thiab WAL-E.
Tsis tas li ntawd, muaj cov xwm txheej uas WAL-G tsis haum rau koj tam sim no, tab sis peb yuav kho nws.
Ua ntej, peb tam sim no tsis muaj built-in backup pov thawj. Peb tsis muaj cov ntaub ntawv pov thawj thaum lub sijhawm thaub qab lossis rov qab. Tau kawg, qhov no yog siv hauv huab. Tab sis qhov no tsuas yog siv los ntawm kev kuaj xyuas ua ntej, tsuas yog los ntawm kev rov ua kom cov pawg. Kuv xav muab qhov no functionality rau cov neeg siv. Tab sis los ntawm kev txheeb xyuas, kuv xav tias hauv WAL-G nws yuav muaj peev xwm rov qab tau cov pawg thiab pib nws, thiab khiav cov kev ntsuam xyuas pa luam yeeb: pg_dumpall rau /dev/null thiab amcheck index verification.
Tam sim no hauv WAL-G tsis muaj txoj hauv kev los ncua ib qho thaub qab ntawm WAL. Ntawd yog, peb txhawb qee qhov rais. Piv txwv li, khaws cia xya hnub dhau los, khaws kaum lub thaub qab kawg, khaws peb lub thaub qab tag nrho. Feem ntau cov neeg tuaj thiab hais tias: "Peb xav tau qhov thaub qab ntawm qhov tshwm sim ntawm Xyoo Tshiab thiab peb xav khaws nws mus ib txhis." WAL-G tseem tsis tau paub yuav ua li cas. (Ceeb Toom - Qhov no twb tau kho lawm. Nyeem ntxiv - Backup-kos xaiv hauv )
Thiab peb tsis muaj nplooj ntawv txheeb xyuas thiab kev kuaj xyuas kev ncaj ncees rau txhua ntu ntu thaum siv PITR.
Los ntawm tag nrho cov no kuv muab tso ua ke ib qhov project rau Google Lub Caij Ntuj Sov ntawm Code. Yog tias koj paub cov tub ntxhais kawm txawj ntse uas xav sau ib yam dab tsi hauv Go thiab tau txais ntau txhiab nyiaj los ntawm ib lub tuam txhab nrog tsab ntawv "G", ces qhia peb qhov project rau lawv. Kuv yuav ua tus cob qhia rau qhov project no, lawv tuaj yeem ua tau. Yog tias tsis muaj cov tub ntxhais kawm, kuv yuav coj nws thiab ua nws tus kheej thaum lub caij ntuj sov.
Thiab peb muaj ntau yam teeb meem me me uas peb maj mam ua haujlwm. Thiab qee qhov zoo nkauj txawv txawv tshwm sim.
Piv txwv li, yog tias koj muab WAL-G ib qho thaub qab khoob, nws tsuas yog poob. Piv txwv li, yog tias koj qhia nws tias nws yuav tsum tau thaub qab ib daim nplaub tshev khoob. Cov ntaub ntawv pg_control yuav tsis nyob ntawd. Thiab nws yuav xav tias nws tsis to taub ib yam dab tsi. Hauv txoj kev xav, qhov no koj yuav tsum tau sau cov lus ib txwm rau tus neeg siv los piav qhia rau nws siv lub cuab yeej. Tab sis qhov no tsis yog ib qho feature ntawm programming, tab sis ib tug feature ntawm ib tug zoo, intelligible lus.
Peb tsis paub yuav ua li cas thaub qab offline. Yog hais tias lub database dag, peb yuav tsis backup nws. Tab sis txhua yam yog yooj yim heev ntawm no. Peb hu thaub qab los ntawm LSN thaum nws pib. LSN ntawm lub hauv paus yuav tsum tau nyeem los ntawm cov ntaub ntawv tswj. Thiab qhov no yog xws li ib tug unrealized feature. Muaj ntau lub tshuab thaub qab tuaj yeem khaws cov ntaub ntawv hauv qab. Thiab nws yooj yim.
Tam sim no peb tsis tuaj yeem daws qhov tsis muaj qhov chaw thaub qab kom raug. Vim tias peb feem ntau ua haujlwm nrog cov thaub qab loj hauv tsev. Thiab lawv tsis tau nyob ib ncig ntawm nws. Tab sis yog tias ib tug neeg xav tau kev pab cuam hauv Go tam sim no, ntxiv kev tuav rau qhov tsis raug ntawm qhov chaw rau lub thoob. Kuv mam li saib mus rau hauv qhov kev thov rub.
Thiab qhov tseem ceeb uas peb txhawj xeeb yog tias peb xav tau ntau qhov kev xeem docker kev sib koom ua ke raws li qhov ua tau uas tshawb xyuas ntau yam xwm txheej. Tam sim no peb tsuas yog sim cov xwm txheej yooj yim xwb. Ntawm txhua qhov kev cog lus, tab sis peb xav tshawb xyuas kev cog lus-los-coj tag nrho cov haujlwm uas peb txhawb nqa. Tshwj xeeb, piv txwv li, peb yuav muaj kev txhawb nqa txaus rau PostgreSQL 9.4-9.5. Peb txhawb lawv vim hais tias cov zej zog txhawb PostgreSQL, tab sis peb tsis kuaj xyuas kev cog lus los ntawm kev cog lus kom paub tseeb tias txhua yam tsis tawg. Thiab nws zoo li kuv tias qhov no yog qhov txaus ntshai heev.
Peb muaj WAL-G khiav ntawm ntau tshaj ib txhiab pawg hauv Yandex Database tswj. Thiab nws thaub qab ntau pua terabytes ntawm cov ntaub ntawv txhua hnub.
Peb muaj TODO ntau hauv peb cov cai. Yog tias koj xav tau kev pab cuam, tuaj, peb tab tom tos rub cov lus thov, peb tos cov lus nug.
Lus nug
Nyob zoo os! Ua tsaug! Kuv kwv yees yog tias yog tias koj siv WAL-delta, tej zaum koj yuav vam khom ntau ntawm nplooj ntawv sau. Thiab yog tias muaj, koj puas tau xeem? Koj tau pom daim duab zoo nkauj. Yuav ua li cas zoo nkauj dua yog tias FPW raug muab tua?
Cov nplooj ntawv sau tag nrho yog qhib rau peb, peb tsis tau sim ua kom tsis raug. Ntawd yog, kuv, ua tus tsim tawm, tsis tau sim tua nws. Cov thawj tswj hwm uas tau tshawb fawb tau tej zaum tshawb fawb txog qhov teeb meem no. Tab sis peb xav tau FPW. Yuav luag tsis muaj leej twg lov tes taw nws, vim hais tias yog tsis li ntawd nws yog tsis yooj yim sua kom coj ib tug thaub qab los ntawm ib tug replica.
Ua tsaug rau daim ntawv qhia! Kuv muaj ob lo lus nug. Thawj lo lus nug yog yuav ua li cas rau tablespaces?
Peb tab tom tos qhov kev thov rub. Peb cov databases nyob ntawm SSD thiab NMVE disks thiab peb tsis xav tau qhov no tiag. Kuv tsis tau npaj siv sijhawm loj los ua kom nws zoo tam sim no. Kuv tag nrho siab ntsws tawm tswv yim tias peb txhawb qhov no. Muaj cov neeg uas txhawb nqa nws, tab sis txhawb nws txoj hauv kev uas haum rau lawv. Lawv ua ib rab diav rawg, tab sis lawv tsis ua rub thov. (Ntxiv rau hauv version 0.2.13)
Thiab lo lus nug thib ob. Koj tau hais thaum pib tias WAL-G xav tias nws ua haujlwm ib leeg thiab tsis xav tau wrappers. Kuv siv wrappers kuv tus kheej. Vim li cas lawv yuav tsum tsis txhob siv?
Peb xav kom nws yooj yim li balalaika. Qhov no txhais tau tias koj tsis xav tau ib yam dab tsi tsuas yog balalaika. Peb xav kom lub system yooj yim. Yog tias koj muaj kev ua haujlwm uas koj yuav tsum tau ua hauv tsab ntawv, ces tuaj qhia peb - ββpeb yuav ua nws hauv Go.
Nyob zoo os! Ua tsaug rau daim ntawv qhia! Peb tsis tuaj yeem tau txais WAL-G los ua haujlwm nrog GPG decryption. Nws encrypts ib txwm, tab sis tsis xav decrypt. Puas yog ib qho uas tsis ua haujlwm rau peb? Qhov xwm txheej yog kev nyuaj siab.
Tsim ib qho teeb meem ntawm GitHub thiab cia peb xam nws tawm.
Qhov ntawd yog, koj tsis tau ntsib qhov no?
Muaj qhov tshwj xeeb ntawm daim ntawv qhia txog kev ua yuam kev uas thaum WAL-G tsis nkag siab tias nws yog cov ntaub ntawv dab tsi, nws nug: "Tej zaum nws yog encrypted?" Tej zaum qhov teeb meem tsis yog encryption txhua. Kuv xav txhim kho qhov kev txiav txim siab ntawm lub ncauj lus no. Nws yuav tsum decipher nws. Tam sim no peb tab tom ua haujlwm ntawm lub ncauj lus no hauv qhov kev nkag siab tias peb tsis nyiam qhov system kom tau txais cov yuam sij rau pej xeem thiab ntiag tug. Vim peb hu rau sab nraud GPG kom nws muab peb cov yuam sij. Thiab tom qab ntawd peb muab cov yuam sij no thiab hloov mus rau sab hauv GPG, uas yog qhib PGP, uas tau muab tso ua ke rau peb hauv WAL-G, thiab qhov ntawd peb hu ua encryption. Hauv qhov no, peb xav txhim kho lub kaw lus thiab xav txhawb Libsodium encryption (Ntxiv rau hauv version 0.2.15). Tau kawg, kev txiav txim siab yuav tsum ua haujlwm, cia peb txheeb xyuas nws - koj xav tau cov tsos mob ntau dua li ob peb lo lus. Koj tuaj yeem sib sau ua ke hauv chav hais lus qee zaum thiab saib lub kaw lus. (PGP encryption yam tsis muaj GPG sab nraud - v0.2.9)
Nyob zoo! Ua tsaug rau daim ntawv qhia! Kuv muaj ob lo lus nug. Kuv muaj lub siab xav coj txawv txawv ua pg_basebackup thiab WAL nkag rau hauv ob lub chaw pabcuam, piv txwv li kuv xav ua ib huab thiab lwm qhov. Puas muaj ib txoj hauv kev los ua qhov no?
Qhov no tsis muaj nyob tam sim no, tab sis nws yog ib lub tswv yim nthuav.
Kuv tsuas yog tsis ntseeg ib tus neeg muab kev pabcuam, kuv xav kom muaj tib yam hauv lwm tus, tsuas yog hauv rooj plaub.
Lub tswv yim yog nthuav. Technically, qhov no tsis yog qhov nyuaj rau kev siv. Txhawm rau tiv thaiv lub tswv yim los ntawm kev ploj, kuv tuaj yeem thov koj ua qhov teeb meem ntawm GitHub?
Yog, tau kawg.
Thiab tom qab ntawd, thaum cov tub ntxhais kawm tuaj rau Google Lub Caij Ntuj Sov ntawm Code, peb yuav ntxiv lawv rau qhov project kom muaj ntau txoj haujlwm kom tau txais ntau dua ntawm lawv.
Thiab lo lus nug thib ob. Muaj qhov teeb meem ntawm GitHub. kuv xav tias twb kaw lawm. Muaj kev ceeb ntshai thaum rov qab los. Thiab kom kov yeej nws, koj tau ua ib lub rooj sib txoos cais. Nws yog txoj cai hauv cov teeb meem. Thiab muaj kev xaiv los ua ib qho chaw sib txawv hauv ib qho xov. Thiab yog vim li cas nws ua haujlwm qeeb heev. Thiab peb tau ntsib qhov teeb meem no, thiab nws tseem tsis tau kho.
Qhov teeb meem yog vim li cas rau qee qhov chaw cia (CEPH) rov pib dua qhov kev sib txuas thaum peb tuaj rau nws nrog siab concurrency. Yuav ua li cas txog qhov no? Lub logic retry zoo li no. Peb tab tom sim rub tawm cov ntaub ntawv dua. Hauv ib kis, peb muaj ntau cov ntaub ntawv tsis rub tawm, peb yuav ua qhov thib ob rau txhua tus uas tsis tau nkag mus. Thiab tsuav yog tsawg kawg ib cov ntaub ntawv yog loaded ib iteration, peb rov ua dua thiab rov ua dua. Peb tau txhim kho lub logic ntawm kev rov ua dua - exponential backoff. Tab sis nws tsis paub meej tias yuav ua li cas nrog qhov tseeb tias qhov kev sib txuas tsuas yog tawg ntawm lub kaw lus cia. Ntawd yog, thaum peb upload rau ib qho kwj, nws tsis tawg cov kev sib txuas. Peb tuaj yeem txhim kho dab tsi ntawm no? Peb muaj network throttling, peb tuaj yeem txwv txhua qhov kev sib txuas los ntawm tus lej ntawm bytes nws xa. Txwv tsis pub, kuv tsis paub yuav ua li cas nrog qhov tseeb tias cov khoom khaws cia tsis tso cai rau peb rub tawm lossis rub tawm los ntawm nws ua ke.
Tsis muaj SLA? Tsis yog sau rau lawv li cas lawv cia lawv tus kheej raug tsim txom?
Lub ntsiab lus yog tias cov neeg uas tuaj nrog lo lus nug no feem ntau muaj lawv lub tsev khaws puav pheej. Ntawd yog, tsis muaj leej twg los ntawm Amazon lossis Google Huab, lossis Yandex Object Storage.
Tej zaum cov lus nug tsis yog rau koj lawm?
Lo lus nug no nyob rau hauv cov ntaub ntawv no tsis muaj teeb meem rau leej twg. Yog tias muaj cov tswv yim hais txog yuav ua li cas nrog qhov no, cia peb ua nws hauv WAL-G. Tab sis txog tam sim no kuv tsis muaj tswv yim zoo txog yuav ua li cas nrog qhov no. Muaj qee qhov Chaw Cia Khoom uas txhawb nqa cov npe thaub qab sib txawv. Koj hais kom lawv sau cov khoom, thiab lawv ntxiv cov ntawv tais ceev tseg rau ntawd. WAL-G ntshai ntawm qhov no - muaj qee yam ntawm no uas tsis yog cov ntaub ntawv, Kuv tuaj yeem rov qab tsis tau, uas txhais tau hais tias qhov thaub qab tsis rov qab los. Ntawd yog, qhov tseeb, koj muaj pawg rov qab ua tiav, tab sis nws xa rov qab rau koj qhov xwm txheej tsis raug vim tias Cov Khoom Cia tau xa rov qab qee cov ntaub ntawv coj txawv txawv uas nws tsis nkag siab tag nrho.
Qhov no yog ib yam uas tshwm sim hauv Mail huab.
Yog tias koj tuaj yeem tsim kev tsim tawm ...
Nws yog reproduced tas li ...
Yog tias muaj kev tsim tawm, ces kuv xav tias peb yuav sim nrog cov tswv yim rov ua dua thiab txheeb xyuas seb yuav rov ua li cas thiab nkag siab tias huab xav tau dab tsi ntawm peb. Tej zaum nws yuav ruaj khov rau peb ntawm peb qhov kev sib txuas thiab yuav tsis tawg qhov kev sib txuas, ces peb yuav ua tib zoo ncav cuag peb. Vim tias tam sim no peb poob qhov kev sib txuas sai heev, piv txwv li yog tias peb pib rov qab nrog 16 threads, tom qab thawj zaug rov ua dua yuav muaj 8 threads, 4 threads, 2 threads thiab ib qho. Thiab ces nws yuav rub cov ntaub ntawv mus rau hauv ib tug kwj. Yog tias muaj qee qhov txiaj ntsig zoo li 7,5 xov yog qhov zoo tshaj plaws rau cov twj tso kua mis, ces peb yuav nyob twj ywm rau lawv thiab sim ua lwm 7,5 xov. Nov yog ib lub tswv yim.
Ua tsaug rau daim ntawv qhia! Kev ua haujlwm tiav rau kev ua haujlwm nrog WAL-G zoo li cas? Piv txwv li, nyob rau hauv cov ntaub ntawv ruam thaum tsis muaj delta hla nplooj ntawv. Thiab peb muab thiab tshem tawm thawj zaug thaub qab, ces archive lub ncej kom txog thaum peb xiav nyob rau hauv lub ntsej muag. Ntawm no, raws li kuv nkag siab nws, muaj kev tawg. Qee lub sij hawm koj yuav tsum tau ua delta thaub qab ntawm nplooj ntawv, piv txwv li qee cov txheej txheem sab nraud yog tsav qhov no lossis qhov no tshwm sim li cas?
delta backup API yog qhov yooj yim heev. Muaj tus lej nyob ntawd - max delta cov kauj ruam, uas yog qhov nws hu ua. Nws defaults rau xoom. Qhov no txhais tau tias txhua zaus koj ua thaub qab-thawj, nws rub tawm tag nrho cov thaub qab. Yog tias koj hloov nws mus rau ib tus lej zoo, piv txwv li, 3, tom qab ntawd lub sijhawm tom ntej koj ua qhov thaub qab-thawj, nws saib keeb kwm ntawm kev thaub qab yav dhau los. Nws pom tias koj tsis tshaj cov saw ntawm 3 deltas thiab ua ib tug delta.
Ntawd yog, txhua zaus peb tso WAL-G, nws sim ua kom tag nrho thaub qab?
Tsis yog, peb khiav WAL-G, thiab nws sim ua lub delta yog tias koj txoj cai tso cai rau nws.
Hais lus ntxhib, yog tias koj khiav nws nrog xoom txhua zaus, nws puas yuav coj zoo li pg_basebackup?
Tsis yog, nws tseem yuav khiav nrawm dua vim nws siv compression thiab parallelism. Pg_basebackup yuav muab cov ncej nyob ib sab ntawm koj. WAL-G xav tias koj tau archiving configured. Thiab nws yuav ceeb toom yog tias nws tsis tau teeb tsa.
Pg_basebackup tuaj yeem khiav yam tsis muaj shafts.
Yog lawm, ces lawv yuav coj zoo li qub. Pg_basebackup luam rau cov ntaub ntawv kaw lus. Los ntawm txoj kev, peb muaj ib tug tshiab feature uas kuv tsis nco qab hais. Tam sim no peb tuaj yeem thaub qab rau cov ntaub ntawv kaw lus los ntawm pg_basebackup. Kuv tsis paub vim li cas qhov no xav tau, tab sis nws muaj.
Piv txwv li, ntawm CephFS. Tsis yog txhua leej txhua tus xav teeb tsa Cov Khoom Cia.
Yog lawm, tej zaum yog vim li cas lawv nug ib lo lus nug txog qhov no kom peb ua tau. Thiab peb tau ua nws.
Ua tsaug rau daim ntawv qhia! Tsuas muaj lus nug txog kev theej rau hauv cov ntaub ntawv kaw lus. Tawm ntawm lub thawv, tam sim no koj puas txhawb nqa kev luam rau cov chaw taws teeb cia, piv txwv li, yog tias muaj qee lub txee rau hauv cov ntaub ntawv chaw lossis lwm yam?
Nyob rau hauv daim ntawv no, qhov no yog ib lo lus nug nyuaj. Yog lawm, peb txhawb nqa, tab sis qhov kev ua haujlwm no tsis suav nrog hauv ib qho kev tso tawm. Ntawd yog, txhua qhov kev tso tawm ua ntej txhawb nqa qhov no, tab sis cov ntawv tso tawm tsis ua. Qhov kev ua haujlwm no tau ntxiv rau hauv version 0.2. Nws yuav twv yuav raug tso tawm sai sai, sai li sai tau thaum peb kho tag nrho cov kab paub. Tab sis tam sim no qhov no tsuas yog ua tiav hauv kev tso tawm ua ntej. Muaj ob kab hauv qhov kev tso tawm ua ntej. Teeb meem nrog WAL-E rov qab, peb tsis tau kho nws. Thiab nyob rau hauv qhov tseeb pre-tso tawm cov kab mob txog delta-backup tau ntxiv. Yog li ntawd, peb xav kom txhua tus siv cov kev tso tawm versions. Thaum tsis muaj kab ntau ntxiv hauv qhov kev tso tawm ua ntej, peb tuaj yeem hais tias peb txhawb nqa Google Cloud, S3-tshaj yam khoom thiab cov ntaub ntawv khaws cia.
Nyob zoo, ua tsaug rau daim ntawv qhia. Raws li kuv nkag siab nws, WAL-G tsis yog qee yam ntawm lub hauv nruab nrab zoo li barmen? Koj puas npaj txav mus rau qhov kev taw qhia no?
Qhov teeb meem yog tias peb tau txav deb ntawm qhov kev taw qhia no. WAL-G nyob ntawm tus tswv tsev, ntawm pawg tswv tsev, thiab ntawm txhua tus tswv hauv pawg. Thaum peb tsiv mus rau ntau txhiab pawg, peb muaj ntau qhov kev sib tw bartender. Thiab txhua zaus ib yam dab tsi poob rau hauv lawv, nws yog ib qho teeb meem loj. Vim tias lawv yuav tsum tau kho, koj yuav tsum nkag siab tias pawg twg tam sim no tsis muaj thaub qab. Kuv tsis npaj los tsim WAL-G hauv kev taw qhia ntawm lub cev kho vajtse rau cov txheej txheem thaub qab. Yog hais tias lub zej zog xav tau ib co functionality ntawm no, kuv tsis mloog kiag li.
Peb muaj pab pawg uas muaj lub luag haujlwm rau kev khaws cia. Thiab peb zoo siab tias nws tsis yog peb, tias muaj cov neeg tshwj xeeb uas muab peb cov ntaub ntawv uas cov ntaub ntawv muaj kev nyab xeeb. Lawv ua txhua yam kev txawj ntse coding nyob rau ntawd los tiv thaiv qhov poob ntawm qee cov ntaub ntawv. Lawv yog lub luag hauj lwm rau network bandwidth. Thaum koj muaj ib tug bartender, koj mam li nco dheev pom tias cov ntaub ntawv me me uas muaj ntau lub tsheb tau sib sau ua ke ntawm tib lub server. Koj zoo li muaj ntau qhov chaw ntawm nws, tab sis rau qee qhov laj thawj txhua yam tsis haum hauv lub network. Nws tuaj yeem tig tawm mus rau lwm txoj hauv kev. Muaj ntau lub network nyob rau ntawd, muaj processor cores, tab sis tsis muaj disks ntawm no. Thiab peb tau nkees ntawm qhov xav tau juggle ib yam dab tsi, thiab peb tsiv mus rau qhov tseeb tias cov ntaub ntawv khaws cia yog ib qho kev pabcuam cais, uas cais cov neeg tshwj xeeb yog lub luag haujlwm.
PS Ib tug tshiab version tau raug tso tawm , nyob rau hauv uas koj tuaj yeem siv .walg.json configuration file, uas nyob rau hauv postgres home directory los ntawm lub neej ntawd. Koj tuaj yeem tso tseg bash scripts. Piv txwv .walg.json nyob rau hauv qhov teeb meem no
Yees duab:
Tau qhov twg los: www.hab.com