RAID 6 plus a hotspare: overkill or just being extra cautious?

文章推薦指數: 80 %
投票人數:10人

Once you get to 2TB or larger drives, you definitely should go with RAID 6, because of the amount of time it can take to recover a drive. Once you're exceeded 9 ... Foundtheinternet!17RAID6plusahotspare:overkillorjustbeingextracautious? Close17Postedby10yearsagoRAID6plusahotspare:overkillorjustbeingextracautious? Iamsettingupanewserver.IamsettingupthediskarrayasRAID6plusahotspare.IknowthatRAID6canloseuptotwodiskswithoutlossofdata.Iwouldratherknowthatifadiskfailsthatthearraywillstartbuildingareplacementwiththehotsparerightaway.Or...isRAID6goodenoughthatthereisnoneedofreservingahotspareanymore?85commentssharesavehidereport80%UpvotedLoginorsignuptoleaveacommentLogInSignUpSortby:best level1·10yr.agoCanneverbetoocareful.26ReplyShareReportSaveFollowlevel1·10yr.agoPrincipleSRE24yrs/13jobsRAID!=backups,nomatterhowyouallocatethedisks.I'dleavewellenoughalonewithRAID6andstartworkonbackupanddisasterrecoveryprocesses.14ReplyShareReportSaveFollowlevel2·10yr.agoTheOPneversaidthatbut,whilewe'reonthesubject:REPLICATION!=backups!7ReplyShareReportSaveFollowContinuethisthread level2Commentdeletedbyuser·10yr.agoContinuethisthread level2·10yr.agoThisisoneofthetimestheOPdidn'tsaythat2ReplyShareReportSaveFollowContinuethisthread level1·10yr.agoWejustlostaSANthatwasconfiguredRAID50withhotspares.Welostadrive,then15hourslaterwiththerebuildat92%,welostaseconddrivefromthesameRAID5.ThiswastheSANthathousedmostofourvmwareenvironment,solosingthearraymeantdowntimeon~75%ofthedatacenter.Obviouslyyouhavetoevaluateyourownsituation,butIwoulddefinitelynotconsiderRAID6plushotsparetobeoverkill.5ReplyShareReportSaveFollowlevel2Op·10yr.agoWhew.Thankyouforconfirmingmyparanoia.3ReplyShareReportSaveFollowlevel2·10yr.agoWhattypeofarray?1ReplyShareReportSaveFollowContinuethisthread level1·10yr.agoJackofAllTradesIruna6x2TBRAID6arrayforpersonalusewitha7th2TBdriveasahotspare.GivesmealittlepeaceofmindincaseIloseadrivewhenI'mnoton-site.3ReplyShareReportSaveFollowlevel1·10yr.agoHowmanydisksinthearray?3ReplyShareReportSaveFollowlevel2Op·10yr.agoSix.1ReplyShareReportSaveFollowContinuethisthread level1·10yr.agoStorageAdminWith7200RPMdrivesover2TBinanysortofquantity,it'snearlyanecessity.3ReplyShareReportSaveFollowlevel2Op·10yr.ago15000RPMdrives,600GBeach,6MbpsSAS.IwasgoingforspeedforaSQLserver.3ReplyShareReportSaveFollowContinuethisthread level1·10yr.agoItdependsonthesizeandnumberofdrives.Ifit'salargearray,thentheanswerisdefinitelyyes.Onceyougetto2TBorlargerdrives,youdefinitelyshouldgowithRAID6,becauseoftheamountoftimeitcantaketorecoveradrive.Onceyou'reexceeded9orsodrives,youshouldstartusingRAID6andahotspare,sincethechanceofadrivefailingcanberelativelyhigh.Ononeserver,Ihave12x2TBdrives,soRAID6withahotsparewastheonlywayIwouldhaveconsideredsettingitup.Youmaynotwanttolosethespaceofonedrive,butthathotsparecanreallysaveyouonalargearray.3ReplyShareReportSaveFollowlevel1·10yr.agoLinuxHPCgraybeardYourquestionleavesoutthemostimportantthing:whatdatawillthisarrayhold?Howimportantisit?Also,whatarethecharacteristicsofdegradedraidsetsinyourraidcontrollers?Istheperformancebadenoughthatitwillbeproblematictoyourusers?What'stherebuildtime?Willaddingorsubtractingthelastdiskchangethestripesizetoasufficientdegreetocauseyourapplicationproblems?Anecdotaldatapoint:atnotimesincebefore2007haveIbeenresponsibleforlessthan800x10-diskRAID6sets(andattimesmanymore),withnospareswhatsoever,andhavesufferedthelossofagrandtotalofzeroraidsets.Weuseratherfancyraidcontrollers,socrappyonboardcardsmaychangethisconsiderably.Also:atacertainpoint,it'saboutwhatmakesyousleepbetter.Ifyou'llsleepbetterwiththespareanditwon'tmaketheperformancesignificantlyworse,youshoulddoit.3ReplyShareReportSaveFollowlevel2·10yr.agoLinuxadmin(HPC)Yeah,wedon'trunhotspares.Butwedoruncrazycount'sofraid61ReplyShareReportSaveFollowlevel1·10yr.agoJackofAllTradesIwouldsaythatitdepends.Ifallthedrivesarefromthesameproductionseries,thentheyareproducedwithinthesametimeperiod.Whichmeansthatwhenonestartsfailing,theotherswillmostlikelyfollowprettywithinashorttimespan.Thisdoesnotneedtobecausedbyproduction,butsimplybadtransportation+++.Note:RebuildingaRAIDwillalsocauseelevatedstressontheotherdrives.Ifthedrivesarefromseparateproductions,thisdangershouldnotbeanythingtoworryabout,andaregularRAID6shouldsuffice.FYI:IhavehadthepleasuretohaveaRAID5lostduetodrivetwofromaseriesdyingwithin24hoursofthefirstdrive,nowidoRAID5;)Thedrivesarefromtwoseries,soisticktonohotspares.2ReplyShareReportSaveFollowlevel1·10yr.agoRAID6+Coldspare(areplacementdrivesomewhereyoucangetitwithin10minutes)willbemorethanenough.Statistically,anyway.2ReplyShareReportSaveFollowlevel2·10yr.agoThisassumesthedatacenter/colocationismanned,andpeoplewillnotice&beabletoreplaceadrivewithin10minutesofafailure.(Thisiswhyhotsparesaresonice)1ReplyShareReportSaveFollowToppostsjune7th2012,Toppostsofjune,2012,Topposts2012,helpRedditcoinsRedditpremiumaboutcareerspressadvertiseblogTermsContentpolicyPrivacypolicyModpolicyRedditInc©2022.AllrightsreservedBacktoTop Advertisement



請為這篇文章評分?