{"id":608,"date":"2011-03-17T11:29:26","date_gmt":"2011-03-17T03:29:26","guid":{"rendered":"http:\/\/www.strongd.net\/?p=608"},"modified":"2011-03-17T11:29:51","modified_gmt":"2011-03-17T03:29:51","slug":"%e5%88%86%e5%b8%83%e5%bc%8f%e6%96%87%e4%bb%b6%e7%b3%bb%e7%bb%9fceph%e8%b0%83%e7%a0%941-%e2%80%93-rados","status":"publish","type":"post","link":"https:\/\/www.strongd.net\/?p=608","title":{"rendered":"\u5206\u5e03\u5f0f\u6587\u4ef6\u7cfb\u7edfCeph\u8c03\u78141 \u2013 RADOS"},"content":{"rendered":"<p>Ceph\u662f\u52a0\u5dde\u5927\u5b66Santa Cruz\u5206\u6821\u7684Sage Weil\uff08DreamHost\u7684\u8054\u5408\u521b\u59cb\u4eba\uff09\u4e13\u4e3a\u535a\u58eb\u8bba\u6587\u8bbe\u8ba1\u7684\u65b0\u4e00\u4ee3\u81ea\u7531\u8f6f\u4ef6\u5206\u5e03\u5f0f\u6587\u4ef6\u7cfb\u7edf\u3002\u81ea2007\u5e74\u6bd5\u4e1a\u4e4b\u540e\uff0cSage\u5f00\u59cb\u5168\u804c\u6295\u5165\u5230Ceph\u5f00 \u53d1\u4e4b\u4e2d\uff0c\u4f7f\u5176\u80fd\u9002\u7528\u4e8e\u751f\u4ea7\u73af\u5883\u3002Ceph\u7684\u4e3b\u8981\u76ee\u6807\u662f\u8bbe\u8ba1\u6210\u57fa\u4e8ePOSIX\u7684\u6ca1\u6709\u5355\u70b9\u6545\u969c\u7684\u5206\u5e03\u5f0f\u6587\u4ef6\u7cfb\u7edf\uff0c\u4f7f\u6570\u636e\u80fd\u5bb9\u9519\u548c\u65e0\u7f1d\u7684\u590d\u5236\u30022010\u5e743 \u6708\uff0cLinus Torvalds\u5c06Ceph client\u5408\u5e76\u5230\u5185 \u68382.6.34\u4e2d\u3002<\/p>\n<p>Ceph\u4e2d\u6709\u5f88\u591a\u5728\u5206\u5e03\u5f0f\u7cfb\u7edf\u9886\u57df\u975e\u5e38\u65b0\u9896\u7684\u6280\u672f\u70b9\uff0c\u5bf9\u89e3\u51b3\u5206\u5e03\u5f0f\u6587\u4ef6\u7cfb\u7edf\u4e2d\u4e00\u4e9b\u5e38\u89c1\u7684\u95ee\u9898\u7684\u7814\u7a76\u975e\u5e38\u6709\u6307\u5bfc\u610f\u4e49\u3002\u6240\u4ee5\u503c\u5f97\u7814\u7a76\u3002<\/p>\n<h1>RADOS\u7b80\u4ecb<\/h1>\n<h2>1 RADOS\u6982\u8ff0<\/h2>\n<p>RADOS (Reliable, Autonomic Distributed Object Store) \u662fCeph\u7684\u6838\u5fc3\u4e4b\u4e00\uff0c\u4f5c\u4e3aCeph\u5206\u5e03\u5f0f\u6587\u4ef6\u7cfb\u7edf\u7684\u4e00\u4e2a\u5b50\u9879\u76ee\uff0c\u7279\u522b\u4e3aCeph\u7684\u9700\u6c42\u8bbe\u8ba1\uff0c\u80fd\u591f\u5728\u52a8\u6001\u53d8\u5316\u548c\u5f02\u8d28\u7ed3\u6784\u7684\u5b58\u50a8\u8bbe\u5907\u673a\u7fa4\u4e4b\u4e0a\u63d0\u4f9b\u4e00\u79cd\u7a33\u5b9a\u3001\u53ef\u6269\u5c55\u3001\u9ad8\u6027\u80fd\u7684\u5355\u4e00\u903b\u8f91\u5bf9\u8c61(Object)\u5b58\u50a8\u63a5\u53e3\u548c\u80fd\u591f\u5b9e\u73b0\u8282\u70b9\u7684\u81ea\u9002\u5e94\u548c\u81ea\u7ba1\u7406\u7684\u5b58\u50a8\u7cfb\u7edf\u3002\u4e8b\u5b9e\u4e0a\uff0cRADOS\u4e5f\u53ef\u4ee5\u5355\u72ec\u4f5c\u4e3a\u4e00\u79cd\u5206\u5e03\u5f0f\u6570\u636e\u5b58\u50a8\u7cfb\u7edf\uff0c\u7ed9\u9002\u5408\u76f8\u5e94\u9700\u6c42\u7684\u5206\u5e03\u5f0f\u6587\u4ef6\u7cfb\u7edf\u63d0\u4f9b\u6570\u636e\u5b58\u50a8\u670d\u52a1\u3002<\/p>\n<h2>2 RADOS\u67b6\u6784\u7b80\u4ecb<\/h2>\n<p>RADOS\u7cfb\u7edf\u4e3b\u8981\u7531\u4e24\u4e2a\u90e8\u5206\u7ec4\u6210(\u5982\u56fe1\u6240\u793a)\uff1a<\/p>\n<p>1\uff0e\u7531\u6570\u76ee\u53ef\u53d8\u7684\u5927\u89c4\u6a21OSDs\uff08Object Storage Devices\uff09\u7ec4\u6210\u7684\u673a\u7fa4\uff0c\u8d1f\u8d23\u5b58\u50a8\u6240\u6709\u7684Objects\u6570\u636e\uff1b<\/p>\n<p>2\uff0e\u7531\u5c11\u91cfMonitors\u7ec4\u6210\u7684\u5f3a\u8026\u5408\u3001\u5c0f\u89c4\u6a21\u673a\u7fa4\uff0c\u8d1f\u8d23\u7ba1\u7406Cluster Map\uff0c\u5176\u4e2dCluster Map\u662f\u6574\u4e2aRADOS\u7cfb\u7edf\u7684\u5173\u952e\u6570\u636e\u7ed3\u6784\uff0c\u7ba1\u7406\u673a\u7fa4\u4e2d\u7684\u6240\u6709\u6210\u5458\u3001\u5173\u7cfb\u3001\u5c5e\u6027\u7b49\u4fe1\u606f\u4ee5\u53ca\u6570\u636e\u7684\u5206\u53d1\u3002<\/p>\n<p><a href=\"http:\/\/www.tbdata.org\/wp-content\/uploads\/2011\/03\/ceph-rados-schema.jpg\"><img loading=\"lazy\" decoding=\"async\" src=\"http:\/\/www.tbdata.org\/wp-content\/uploads\/2011\/03\/ceph-rados-schema.jpg\" alt=\"\" width=\"455\" height=\"263\" \/><\/a><\/p>\n<p>\u56fe1 RADOS\u7cfb\u7edf\u67b6\u6784\u56fe\u793a<\/p>\n<p>\u5bf9\u4e8eRADOS\u7cfb\u7edf\uff0c\u8282\u70b9\u7ec4\u7ec7\u7ba1\u7406\u548c\u6570\u636e\u5206\u53d1\u7b56\u7565\u5747\u6709\u5185\u90e8\u7684Monitors\u5168\u6743\u8d1f\u8d23\uff0c\u6240\u4ee5\uff0c\u4eceClients\u89d2\u5ea6\u8bbe\u8ba1\u76f8\u5bf9\u6bd4\u8f83\u7b80\u5355\uff0c\u5b83\u7ed9\u5e94\u7528\u63d0\u4f9b\u7684\u4ec5\u4e3a\u7b80\u5355\u7684\u5b58\u50a8\u63a5\u53e3\u3002<\/p>\n<h2>3 RADOS\u8be6\u7ec6\u4ecb\u7ecd<\/h2>\n<h3>3.1 \u6269\u5c55\u673a\u7fa4<\/h3>\n<p><strong>1<\/strong><strong>\uff0eCluster Map<\/strong><\/p>\n<p>\u5b58\u50a8\u673a\u7fa4\u7684\u7ba1\u7406\uff0c\u552f\u4e00\u7684\u9014\u5f84\u662fCluster Map\u901a\u8fc7\u5bf9Monitor Cluster\u64cd\u4f5c\u5b8c\u6210\u3002Cluster Map\u662f\u6574\u4e2aRADOS\u7cfb\u7edf\u7684\u6838\u5fc3\u6570\u636e\u7ed3\u6784\uff0c\u5176\u4e2d\u6307\u5b9a\u4e86\u673a\u7fa4\u4e2d\u7684OSDs\u4fe1\u606f\u548c\u6240\u6709\u6570\u636e\u7684\u5206\u5e03\u60c5\u51b5\u3002\u6240\u6709\u6d89\u53ca\u5230RADOS\u7cfb\u7edf\u7684Storage\u8282\u70b9\u548cClients\u90fd\u6709\u6700\u65b0epoch\u7684Cluster Map\u526f\u672c\u3002\u56e0\u4e3aCluster Map\u7684\u7279\u6b8a\u6027\uff0cClient\u5411\u4e0a\u63d0\u4f9b\u4e86\u975e\u5e38\u7b80\u5355\u7684\u63a5\u53e3\u5b9e\u73b0\u5c06\u6574\u4e2a\u5b58\u50a8\u673a\u7fa4\u62bd\u8c61\u4e3a\u5355\u4e00\u7684\u903b\u8f91\u5bf9\u8c61\u5b58\u50a8\u7ed3\u6784\u3002<\/p>\n<p>Cluster Map\u7684\u66f4\u65b0\u7531OSD\u7684\u72b6\u6001\u53d8\u5316\u6216\u8005\u5176\u4ed6\u4e8b\u4ef6\u9020\u6210\u6570\u636e\u5c42\u7684\u53d8\u5316\u9a71\u52a8\uff0c\u6bcf\u4e00\u6b21Cluster Map\u66f4\u65b0\u90fd\u9700\u8981\u5c06map epoch\u589e\u52a0\uff0cmap epoch\u4f7fCluster Map\u5728\u6240\u6709\u8282\u70b9\u4e0a\u7684\u526f\u672c\u90fd\u4fdd\u6301\u540c\u6b65\uff0c\u540c\u65f6\uff0cmap epoch\u53ef\u4ee5\u4f7f\u4e00\u4e9b\u8fc7\u671f\u7684Cluster Map\u80fd\u591f\u901a\u8fc7\u901a\u4fe1\u5bf9\u7b49\u8282\u70b9\u53ca\u65f6\u66f4\u65b0\u3002<\/p>\n<p>\u5728\u5927\u89c4\u6a21\u7684\u5206\u5e03\u5f0f\u7cfb\u7edf\u4e2d\uff0cOSDs\u7684failures\/recoveries\u662f\u5e38\u89c1\u7684\uff0c\u6240\u4ee5\uff0cCluster Map\u7684\u66f4\u65b0\u5c31\u6bd4\u8f83\u9891\u7e41\uff0c\u5982\u679c\u5c06\u6574\u4e2aCluster Map\u8fdb\u884c\u5206\u53d1\u6216\u5e7f\u64ad\u663e\u7136\u4f1a\u9020\u6210\u8d44\u6e90\u7684\u6d6a\u8d39\uff0cRADOS\u91c7\u7528\u5206\u53d1incremental map\u7684\u7b56\u7565\u907f\u514d\u8d44\u6e90\u6d6a\u8d39\uff0c\u5176\u4e2dincremental map\u4ec5\u5305\u542b\u4e86\u4e24\u4e2a\u8fde\u7eedepoch\u4e4b\u95f4Cluster Map\u7684\u589e\u91cf\u4fe1\u606f\u3002<\/p>\n<p><strong>2<\/strong><strong>\uff0eData Placement<\/strong><\/p>\n<p>\u6570\u636e\u8fc1\u79fb\uff1a\u5f53\u6709\u65b0\u7684\u50a8\u5b58\u8bbe\u5907\u52a0\u5165\u65f6\uff0c\u673a\u7fa4\u4e0a\u7684\u6570\u636e\u4f1a\u968f\u673a\u7684\u9009\u51fa\u4e00\u90e8\u5206\u8fc1\u79fb\u5230\u65b0\u7684\u8bbe\u5907\u4e0a\uff0c\u7ef4\u6301\u73b0\u6709\u5b58\u50a8\u7ed3\u6784\u7684\u5e73\u8861\u3002<\/p>\n<p>\u6570\u636e\u5206\u53d1\uff1a\u901a\u8fc7\u4e24\u4e2a\u9636\u6bb5\u7684\u8ba1\u7b97\u5f97\u5230\u5408\u9002\u7684Object\u7684\u5b58\u50a8\u4f4d\u7f6e\u3002\u5982\u56fe2\u6240\u793a\u3002<\/p>\n<p><a href=\"http:\/\/www.tbdata.org\/wp-content\/uploads\/2011\/03\/ceph-rados-placement.jpg\"><img loading=\"lazy\" decoding=\"async\" src=\"http:\/\/www.tbdata.org\/wp-content\/uploads\/2011\/03\/ceph-rados-placement.jpg\" alt=\"\" width=\"432\" height=\"211\" \/><\/a><\/p>\n<p>\u56fe2 \u6570\u636e\u5206\u53d1\u56fe\u793a<\/p>\n<p>1\uff0eObject\u5230PG\u7684\u6620\u5c04\u3002PG (Placement Group)\u662fObjects\u7684\u903b\u8f91\u96c6\u5408\u3002\u76f8\u540cPG\u91cc\u7684Object\u4f1a\u88ab\u7cfb\u7edf\u5206\u53d1\u5230\u76f8\u540c\u7684OSDs\u96c6\u5408\u4e2d\u3002\u7531Object\u7684\u540d\u79f0\u901a\u8fc7Hash\u7b97\u6cd5\u5f97\u5230\u7684\u7ed3\u679c\u7ed3\u5408\u5176\u4ed6\u4e00\u4e9b\u4fee\u6b63\u53c2\u6570\u53ef\u4ee5\u5f97\u5230Object\u6240\u5bf9\u5e94\u7684PG\u3002<\/p>\n<p>2\uff0eRADOS\u7cfb\u7edf\u6839\u636e\u6839\u636eCluster Map\u5c06PGs\u5206\u914d\u5230\u76f8\u5e94\u7684OSDs\u3002\u8fd9\u7ec4OSDs\u6b63\u662fPG\u4e2d\u7684Objects\u6570\u636e\u7684\u5b58\u50a8\u4f4d\u7f6e\u3002RADOS\u91c7\u7528CRUSH\u7b97\u6cd5\u5b9e\u73b0\u4e86\u4e00\u79cd\u7a33\u5b9a\u3001\u4f2a\u968f\u673a\u7684hash\u7b97\u6cd5\u3002CRUSH\u5b9e\u73b0\u4e86\u5e73\u8861\u7684\u548c\u4e0e\u5bb9\u91cf\u76f8\u5173\u7684\u6570\u636e\u5206\u914d\u7b56\u7565\u3002CRUSH\u5f97\u5230\u7684\u4e00\u7ec4OSDs\u8fd8\u4e0d\u662f\u6700\u7ec8\u7684\u6570\u636e\u5b58\u50a8\u76ee\u6807\uff0c\u9700\u8981\u7ecf\u8fc7\u521d\u6b65\u7684filter\uff0c\u56e0\u4e3a\u5bf9\u4e8e\u5927\u89c4\u6a21\u7684\u5206\u5e03\u5f0f\u673a\u7fa4\uff0c\u5b95\u673a\u7b49\u539f\u56e0\u4f7f\u5f97\u90e8\u5206\u8282\u70b9\u53ef\u80fd\u5931\u6548\uff0cfilter\u5c31\u662f\u4e3a\u8fc7\u6ee4\u8fd9\u4e9b\u8282\u70b9\uff0c\u5982\u679c\u8fc7\u6ee4\u540e\u5b58\u50a8\u76ee\u6807\u4e0d\u80fd\u6ee1\u8db3\u4f7f\u7528\u5219\u963b\u585e\u5f53\u524d\u64cd\u4f5c\u3002<\/p>\n<p><strong>3<\/strong><strong>\uff0eDevice State<\/strong><\/p>\n<p>Cluster Map\u4e2d\u5173\u4e8eDevice State\u7684\u63cf\u8ff0\u89c1\u4e0b\u8868\u6240\u793a\u3002<\/p>\n<p>\u88681 Device State\u63cf\u8ff0<\/p>\n<table border=\"1\" cellspacing=\"0\" cellpadding=\"0\">\n<tbody>\n<tr>\n<td width=\"92\" valign=\"top\"><\/td>\n<td width=\"142\" valign=\"top\">\u2014<\/td>\n<td width=\"208\" valign=\"top\">in<\/td>\n<td width=\"126\" valign=\"top\">out<\/td>\n<\/tr>\n<tr>\n<td width=\"92\" valign=\"top\">\u2014<\/td>\n<td width=\"142\" valign=\"top\">\u2014<\/td>\n<td width=\"208\" valign=\"top\">assigned PGs<\/td>\n<td width=\"126\" valign=\"top\">not assigned PGs<\/td>\n<\/tr>\n<tr>\n<td width=\"92\" valign=\"top\">up<\/td>\n<td width=\"142\" valign=\"top\">online &amp; reachable<\/td>\n<td width=\"208\" valign=\"top\">active<\/td>\n<td width=\"126\" valign=\"top\">online &amp; idle<\/td>\n<\/tr>\n<tr>\n<td width=\"92\" valign=\"top\">down<\/td>\n<td width=\"142\" valign=\"top\">unreachable<\/td>\n<td width=\"208\" valign=\"top\">unreachable &amp; not remapped<\/td>\n<td width=\"126\" valign=\"top\">failed<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<p><strong>4<\/strong><strong>\uff0eMap propagate<\/strong><\/p>\n<p>Cluster Map\u5728OSD\u4e4b\u95f4\u7684\u66f4\u65b0\u662f\u901a\u8fc7\u4e00\u79cd\u62a2\u5360\u5f0f\u7684\u65b9\u6cd5\u8fdb\u884c\u3002Cluster Map epoch\u7684\u5dee\u5f02\u53ea\u6709\u5728\u4e24\u4e2a\u901a\u4fe1\u5b9e\u4f53\u4e4b\u95f4\u6709\u610f\u4e49\uff0c\u4e24\u4e2a\u901a\u4fe1\u5b9e\u4f53\u5728\u8fdb\u884c\u4fe1\u606f\u4ea4\u6362\u4e4b\u524d\u90fd\u9700\u8981\u4ea4\u6362epoch\uff0c\u4fdd\u8bc1Cluster Map\u7684\u540c\u6b65\u3002\u8fd9\u4e00\u5c5e\u6027\u4f7f\u5f97Cluster Map\u5728\u901a\u4fe1\u5b9e\u4f53\u5185\u90e8\u4e4b\u95f4\u7684\u66f4\u65b0\u5206\u62c5\u4e86\u5168\u5c40\u7684Cluster Map\u5206\u53d1\u538b\u529b\u3002<\/p>\n<p>\u6bcf\u4e00\u4e2aOSD\u90fd\u4f1a\u7f13\u5b58\u6700\u8fd1Cluster Map\u548c\u5230\u5f53\u524d\u65f6\u523b\u7684\u6240\u6709incremental map\u4fe1\u606f\uff0cOSD\u7684\u6240\u6709message\u90fd\u4f1a\u5d4c\u5165incremental map\uff0c\u540c\u65f6\u4fa6\u542c\u4e0e\u5176\u901a\u4fe1\u7684peer\u7684Cluster Map epoch\u3002\u5f53\u4ecepeer\u6536\u5230\u7684message\u4e2d\u53d1\u73b0\u5176epoch\u662f\u8fc7\u671f\u7684\uff0cOSD share\u76f8\u5bf9peer\u6765\u8bf4\u7684incremental map\uff0c\u4f7f\u901a\u4fe1\u7684peers\u90fd\u4fdd\u6301\u540c\u6b65\uff1b\u540c\u6837\u7684\uff0c\u5f53\u4ecepeer\u6536\u5230message\u4e2d\u53d1\u73b0\u672c\u5730epoch\u8fc7\u671f\uff0c\u4ece\u5176\u5d4c\u5165\u5230message\u4e2d\u7684incremental map\u4e2d\u5206\u6790\u5f97\u5230\u76f8\u5bf9\u672c\u5730\u7684incremental map\u7136\u540e\u66f4\u65b0\uff0c\u4fdd\u6301\u540c\u6b65\u3002<\/p>\n<p>\u4e0d\u662f\u540c\u4e00\u4e2a\u901a\u4fe1\u5bf9\u7b49\u65b9\u7684\u4e24\u4e2aOSD\u4e4b\u95f4\u7684epoch\u5dee\u5f02\uff0c\u4e0d\u5f71\u54cd\u540c\u6b65\u3002<\/p>\n<h3>3.2 \u667a\u80fd\u5b58\u50a8<\/h3>\n<p><strong>1<\/strong><strong>\uff0e<\/strong><strong>Replication<\/strong><\/p>\n<p>RADOS\u5b9e\u73b0\u4e86\u4e09\u79cd\u4e0d\u540c\u7684Replication\u65b9\u6848\uff0c\u89c1\u4e0b\u56fe3\u793a\uff1a<\/p>\n<p><a href=\"http:\/\/www.tbdata.org\/wp-content\/uploads\/2011\/03\/ceph-rados-replications.jpg\"><img loading=\"lazy\" decoding=\"async\" src=\"http:\/\/www.tbdata.org\/wp-content\/uploads\/2011\/03\/ceph-rados-replications.jpg\" alt=\"\" width=\"482\" height=\"331\" \/><\/a><\/p>\n<p>\u56fe3 RADOS\u5b9e\u73b0\u7684\u4e09\u79cdreplication\u65b9\u6848<\/p>\n<p>Primary-copy\uff1a\u8bfb\u5199\u64cd\u4f5c\u5747\u5728primary OSD\u4e0a\u8fdb\u884c\uff0c\u5e76\u884c\u66f4\u65b0replicas\uff1b<\/p>\n<p>Chain\uff1a\u94fe\u5f0f\u8bfb\u5199\uff0c\u8bfb\u5199\u5206\u79bb\uff1b<\/p>\n<p>Spaly\uff1aPrimary-copy\u548cChain\u7684\u6298\u4e2d\u65b9\u6848\uff1a\u5e76\u884c\u66f4\u65b0replicas\u548c\u8bfb\u5199\u5206\u79bb\u3002<\/p>\n<p><strong>2<\/strong><strong>\uff0e<\/strong><strong>Consistency<\/strong><\/p>\n<p>\u4e00\u81f4\u6027\u95ee\u9898\u4e3b\u8981\u6709\u4e24\u4e2a\u65b9\u9762\uff0c\u5206\u522b\u662fUpdate\u548cRead\uff1a<\/p>\n<ol>\n<li>Update\uff1a\u5728RADOS\u7cfb\u7edf\u4e2d\u6240\u6709Message\u90fd\u5d4c\u5165\u4e86\u53d1\u9001\u7aef\u7684map epoch\u534f\u8c03\u673a\u7fa4\u7684\u4e00\u81f4\u6027\u3002<\/li>\n<li>Read\uff1a\u4e3a\u907f\u514d\u90e8\u5206OSD\u5931\u6548\u5bfc\u81f4\u6570\u636e\u4e0d\u80fd\u4ece\u8be5OSD\u8bfb\u9700\u8981\u8f6c\u5411\u65b0\u7684OSD\uff0c\u4f46\u662fread operation\u7684\u53d1\u8d77\u65b9\u8fd8\u6ca1\u6709\u8be5OSD\u7684\u5931\u6548\u4fe1\u606f\u7684\u95ee\u9898\uff0c\u540c\u4e00\u4e2aPG\u6240\u5728\u7684OSDs\u9700\u8981\u5b9e\u65f6\u4ea4\u6362Heartbeat\u3002<\/li>\n<\/ol>\n<p><strong>3<\/strong><strong>\uff0e<\/strong><strong>Failure Detection<\/strong><\/p>\n<p>\u9519\u8bef\u68c0\u6d4b\uff1aRADOS\u91c7\u53d6\u5f02\u6b65\u3001\u6709\u5e8f\u7684\u70b9\u5bf9\u70b9Heartbeat\u3002(\u6b64\u5904\u7684\u9519\u8bef\u68c0\u6d4b\u662fOSDs\u81ea\u8eab\u68c0\u6d4b)<\/p>\n<p><strong>4<\/strong><strong>\uff0e<\/strong><strong>Data Migration &amp; Failure Recovery<\/strong><\/p>\n<p>\u7531\u4e8e\u8bbe\u5907\u5931\u6548\u3001\u673a\u7fa4\u6269\u5c55\u3001\u9519\u8bef\u6062\u590d\u9020\u6210\u7684Cluster Map\u66f4\u65b0\u4f7f\u5f97PG\u5230OSDs\u7684\u5bf9\u5e94\u5173\u7cfb\u53d1\u751f\u4e86\u53d8\u5316\uff0c\u4e00\u65e6Cluster Map\u53d1\u751f\u53d8\u5316\uff0c\u76f8\u5e94\u7684OSDs\u4e0a\u7684\u6570\u636e\u4e5f\u9700\u8981\u505a\u76f8\u5e94\u7684\u8c03\u6574\u3002<\/p>\n<p>\u6570\u636e\u7684\u79fb\u690d\u548c\u6570\u636e\u6062\u590d\u90fd\u662f\u7531Primary OSD\u8d1f\u8d23\u7edf\u4e00\u5b8c\u6210\u3002<\/p>\n<p>(Data Migration &amp; Failure Recovery\u5177\u4f53\u65b9\u6cd5\u5f85\u7eed)<\/p>\n<h3>3.3 Monitors<\/h3>\n<p>Monitors\u662fCluster Map\u4e3b\u5907\u4efd\u5b58\u50a8\u76ee\u6807\uff0c\u6240\u6709\u5176\u4ed6\u4f4d\u7f6e\u4e0a\u7684Cluster Map\u6700\u521d\u90fd\u662f\u4eceMonitors\u8bf7\u6c42\u5f97\u5230\u3002Monitors\u901a\u8fc7\u5bf9Cluster Map\u7684\u5468\u671f\u66f4\u65b0\u5347\u7ea7\u5b9e\u73b0\u5b58\u50a8\u673a\u7fa4\u7684\u7ba1\u7406\u3002<\/p>\n<p>Monitor\u7684\u5de5\u4f5c\u5206\u4e24\u4e2a\u9636\u6bb5\uff1a<\/p>\n<p>1\uff0e\u9996\u5148\u5728\u591a\u4e2aMonitors\u4e2d\u9009\u4e3eLeader\uff0c\u4e4b\u540eLeader\u5411\u6240\u6709Monitors\u8bf7\u6c42Map Epoch\uff0cMonitors\u5468\u671f\u6027\u5411Leader\u6c47\u62a5\u7ed3\u679c\u5e76\u544a\u77e5\u5176\u6d3b\u8dc3(Active Monitor)\uff0cLeader\u7edf\u8ba1Quorum\u3002\u8fd9\u9636\u6bb5\u7684\u610f\u4e49\u662f\u4fdd\u8bc1\u6240\u6709\u7684Monitors\u7684Map Epoch\u90fd\u662f\u6700\u65b0\u7684\uff0c\u901a\u8fc7Incremental updates\u5bf9\u5df2\u5931\u6548\u7684Cluster Map\u8fdb\u884c\u66f4\u65b0\u3002<\/p>\n<p>2\uff0eLeader\u5468\u671f\u5411\u6bcf\u4e00\u4e2aActive Monitor\u6388\u6743\u8bb8\u53ef\u63d0\u4f9b\u5206\u53d1Cluster Map\u526f\u672c\u7ed9OSDs\u548cClients\u7684\u670d\u52a1\u3002\u5f53\u6388\u6743\u5931\u6548\u4f46Leader\u4ecd\u6ca1\u6709\u91cd\u65b0\u5206\u53d1\u8ba4\u4e3aLeader died\uff0c\u6b64\u65f6\u91cd\u56de\u7b2c\u4e00\u9636\u6bb5\u8fdb\u884cLeader\u91cd\u9009\uff1b\u5f53Active Monitor\u6ca1\u6709\u5468\u671f\u5411Leader\u53cd\u9988ACK\u5219\u8ba4\u4e3a\u6709Monitor died\uff0c\u91cd\u56de\u7b2c\u4e00\u9636\u6bb5\u8fdb\u884cLeader\u9009\u4e3e\u5e76\u66f4\u65b0Quorum\u3002Leader\u5468\u671f\u5206\u53d1Lease\u548cActive Monitor\u5468\u671f\u53cd\u9988ACK\u7684\u53e6\u5916\u4e00\u4e2a\u4f5c\u7528\u662f\u540c\u6b65Monitors\u7684Cluster Map\u3002Active Monitor\u6536\u5230Update\u8bf7\u6c42\u65f6\uff0c\u9996\u5148\u9a8c\u8bc1\u5f53\u524d\u7684Epoch\u662f\u5426\u4e3a\u6700\u65b0\uff0c\u5982\u679c\u4e0d\u662f\uff0c\u66f4\u65b0\u540e\u5411\u4e0a\u6c47\u62a5\u5230Leader\uff0cLeader\u5206\u53d1\u7ed9\u6240\u6709\u7684Monitors\uff0c\u540c\u65f6\u56de\u6536\u6388\u6743\uff0c\u91cd\u65b0\u5f00\u59cb\u65b0\u4e00\u8f6e\u7684Leader\u9009\u4e3e\u5230Cluster Map\u670d\u52a1\u3002<\/p>\n<p>\u901a\u5e38Monitor\u7684\u8d1f\u8f7d\u6bd4\u8f83\u5c0f\uff1aOSDs\u4e0a\u7684Cluster Map\u66f4\u65b0\u901a\u8fc7OSDs\u4e4b\u95f4\u7684\u673a\u5236\u5b9e\u73b0\uff1bOSDs\u7684\u72b6\u6001\u53d8\u5316\u6bd4\u8f83\u7f55\u89c1\u4e0d\u4f1a\u5bf9Monitors\u7684\u8d1f\u8f7d\u9020\u6210\u5f71\u54cd\u3002\u4f46\u662f\u4e00\u4e9b\u7279\u6b8a\u60c5\u51b5\u53ef\u80fd\u4f1a\u5bf9Monitors\u8d1f\u8f7d\u5e26\u6765\u5f71\u54cd\uff0c\u6bd4\u5982\uff1a\u540c\u65f6\u6709n OSDs failed\uff0c\u6bcf\u4e00\u4e2aOSD store m\u4e2aPGs\uff0c\u6b64\u65f6\u4f1a\u5f62\u6210m\u00d7n\u4e2afailure report\u5230\u8fbeMonitors\uff0c\u5bf9\u4e8e\u89c4\u6a21\u8f83\u5927\u7684\u673a\u7fa4\u8fd9\u6837\u7684\u6570\u636e\u91cf\u6bd4\u8f83\u5927\u3002\u4e3a\u907f\u514d\u8fd9\u79cd\u60c5\u51b5\u7ed9Monitor\u5e26\u6765\u7684\u8d1f\u8f7d\u538b\u529b\uff0cOSDs\u91c7\u7528\u4f2a\u968f\u673a\u7684\u65f6\u95f4\u95f4\u9694\u4ea4\u9519\u5b89\u6392failure\u68c0\u6d4b(\u6b64\u5904\u662f\u4eceOSDs\u5230Monitor\u7684\u68c0\u6d4b)\u5411\u4e0a\u6c47\u62a5\uff0c\u53e6\u5916\u6839\u636eMonitors\u7684\u5e76\u884c\u5316\u548c\u8d1f\u8f7d\u5747\u8861\u5206\u914d\u7684\u7279\u70b9\uff0c\u6269\u5c55Monitors\u662f\u89e3\u51b3Monitors\u7684\u8d1f\u8f7d\u538b\u529b\u7684\u53e6\u4e00\u63aa\u65bd\u3002<\/p>\n<h2>4 \u603b\u7ed3<\/h2>\n<p>\u4e0e\u4f20\u7edf\u7684\u5206\u5e03\u5f0f\u6570\u636e\u5b58\u50a8\u4e0d\u540c\uff0cRADOS\u6700\u5927\u7684\u7279\u70b9\u662f\uff1a<\/p>\n<p>1\uff0e\u5c06\u6587\u4ef6\u6620\u5c04\u5230Objects\u540e\u5229\u7528Cluster Map\u901a\u8fc7CRUSH\u8ba1\u7b97\u800c\u4e0d\u662f\u67e5\u627e\u8868\u65b9\u5f0f\u5b9a\u4f4d\u6587\u4ef6\u6570\u636e\u5728\u5b58\u50a8\u8bbe\u5907\u4e2d\u7684\u4f4d\u7f6e\u3002\u7701\u53bb\u4e86\u4f20\u7edf\u7684File\u5230Block\u7684\u6620\u5c04\u548cBlockMap\u7ba1\u7406\u3002<\/p>\n<p>2\uff0eRADOS\u5145\u5206\u5229\u7528\u4e86OSDs\u7684\u667a\u80fd\u7279\u70b9\uff0c\u5c06\u90e8\u5206\u4efb\u52a1\u6388\u6743\u7ed9OSDs\uff0c\u6700\u5927\u7a0b\u5ea6\u7684\u5b9e\u73b0\u53ef\u6269\u5c55\u3002<\/p>\n<h2>5 \u53c2\u8003\u6587\u732e<\/h2>\n<p>[1]\u00a0\u00a0\u00a0\u00a0 RADOS: A Scalable, Reliable Storage Service for Petabyte-scale Storage Clusters.<\/p>\n<p>[2]\u00a0\u00a0\u00a0\u00a0 Ceph: A Scalable, High-Performance Distributed File System.<\/p>\n<p>&nbsp;<\/p>\n","protected":false},"excerpt":{"rendered":"<p>Ceph\u662f\u52a0\u5dde\u5927\u5b66Santa Cruz\u5206\u6821\u7684Sage Weil\uff08DreamHost\u7684\u8054\u5408\u521b\u59cb\u4eba\uff09\u4e13\u4e3a\u535a\u58eb\u8bba\u6587\u8bbe\u8ba1\u7684\u65b0\u4e00\u4ee3\u81ea\u7531\u8f6f\u4ef6\u5206\u5e03\u5f0f\u6587\u4ef6\u7cfb\u7edf\u3002\u81ea2007\u5e74\u6bd5\u4e1a\u4e4b\u540e\uff0cSage\u5f00\u59cb\u5168\u804c\u6295\u5165\u5230Ceph\u5f00 \u53d1\u4e4b\u4e2d\uff0c\u4f7f\u5176\u80fd\u9002\u7528\u4e8e\u751f\u4ea7\u73af\u5883\u3002Ceph\u7684\u4e3b\u8981\u76ee\u6807\u662f\u8bbe\u8ba1\u6210\u57fa\u4e8ePOSIX\u7684\u6ca1\u6709\u5355\u70b9\u6545\u969c\u7684\u5206\u5e03\u5f0f\u6587\u4ef6\u7cfb\u7edf\uff0c\u4f7f\u6570\u636e\u80fd\u5bb9\u9519\u548c\u65e0\u7f1d\u7684\u590d\u5236\u30022010\u5e743 \u6708\uff0cLinus Torvalds\u5c06Ceph client\u5408\u5e76\u5230\u5185 \u68382.6.34\u4e2d\u3002 Ceph\u4e2d\u6709\u5f88\u591a\u5728\u5206\u5e03\u5f0f\u7cfb\u7edf\u9886\u57df\u975e\u5e38\u65b0\u9896\u7684\u6280\u672f\u70b9\uff0c\u5bf9\u89e3\u51b3\u5206\u5e03\u5f0f\u6587\u4ef6\u7cfb\u7edf\u4e2d\u4e00\u4e9b\u5e38\u89c1\u7684\u95ee\u9898\u7684\u7814\u7a76\u975e\u5e38\u6709\u6307\u5bfc\u610f\u4e49\u3002\u6240\u4ee5\u503c\u5f97\u7814\u7a76\u3002 RADOS\u7b80\u4ecb 1 RADOS\u6982\u8ff0 RADOS (Reliable, Autonomic Distributed Object Store) \u662fCeph\u7684\u6838\u5fc3\u4e4b\u4e00\uff0c\u4f5c\u4e3aCeph\u5206\u5e03\u5f0f\u6587\u4ef6\u7cfb\u7edf\u7684\u4e00\u4e2a\u5b50\u9879\u76ee\uff0c\u7279\u522b\u4e3aCeph\u7684\u9700\u6c42\u8bbe\u8ba1\uff0c\u80fd\u591f\u5728\u52a8\u6001\u53d8\u5316\u548c\u5f02\u8d28\u7ed3\u6784\u7684\u5b58\u50a8\u8bbe\u5907\u673a\u7fa4\u4e4b\u4e0a\u63d0\u4f9b\u4e00\u79cd\u7a33\u5b9a\u3001\u53ef\u6269\u5c55\u3001\u9ad8\u6027\u80fd\u7684\u5355\u4e00\u903b\u8f91\u5bf9\u8c61(Object)\u5b58\u50a8\u63a5\u53e3\u548c\u80fd\u591f\u5b9e\u73b0\u8282\u70b9\u7684\u81ea\u9002\u5e94\u548c\u81ea\u7ba1\u7406\u7684\u5b58\u50a8\u7cfb\u7edf\u3002\u4e8b\u5b9e\u4e0a\uff0cRADOS\u4e5f\u53ef\u4ee5\u5355\u72ec\u4f5c\u4e3a\u4e00\u79cd\u5206\u5e03\u5f0f\u6570\u636e\u5b58\u50a8\u7cfb\u7edf\uff0c\u7ed9\u9002\u5408\u76f8\u5e94\u9700\u6c42\u7684\u5206\u5e03\u5f0f\u6587\u4ef6\u7cfb\u7edf\u63d0\u4f9b\u6570\u636e\u5b58\u50a8\u670d\u52a1\u3002 2 RADOS\u67b6\u6784\u7b80\u4ecb RADOS\u7cfb\u7edf\u4e3b\u8981\u7531\u4e24\u4e2a\u90e8\u5206\u7ec4\u6210(\u5982\u56fe1\u6240\u793a)\uff1a 1\uff0e\u7531\u6570\u76ee\u53ef\u53d8\u7684\u5927\u89c4\u6a21OSDs\uff08Object Storage Devices\uff09\u7ec4\u6210\u7684\u673a\u7fa4\uff0c\u8d1f\u8d23\u5b58\u50a8\u6240\u6709\u7684Objects\u6570\u636e\uff1b 2\uff0e\u7531\u5c11\u91cfMonitors\u7ec4\u6210\u7684\u5f3a\u8026\u5408\u3001\u5c0f\u89c4\u6a21\u673a\u7fa4\uff0c\u8d1f\u8d23\u7ba1\u7406Cluster Map\uff0c\u5176\u4e2dCluster Map\u662f\u6574\u4e2aRADOS\u7cfb\u7edf\u7684\u5173\u952e\u6570\u636e\u7ed3\u6784\uff0c\u7ba1\u7406\u673a\u7fa4\u4e2d\u7684\u6240\u6709\u6210\u5458\u3001\u5173\u7cfb\u3001\u5c5e\u6027\u7b49\u4fe1\u606f\u4ee5\u53ca\u6570\u636e\u7684\u5206\u53d1\u3002 \u56fe1 RADOS\u7cfb\u7edf\u67b6\u6784\u56fe\u793a \u5bf9\u4e8eRADOS\u7cfb\u7edf\uff0c\u8282\u70b9\u7ec4\u7ec7\u7ba1\u7406\u548c\u6570\u636e\u5206\u53d1\u7b56\u7565\u5747\u6709\u5185\u90e8\u7684Monitors\u5168\u6743\u8d1f\u8d23\uff0c\u6240\u4ee5\uff0c\u4eceClients\u89d2\u5ea6\u8bbe\u8ba1\u76f8\u5bf9\u6bd4\u8f83\u7b80\u5355\uff0c\u5b83\u7ed9\u5e94\u7528\u63d0\u4f9b\u7684\u4ec5\u4e3a\u7b80\u5355\u7684\u5b58\u50a8\u63a5\u53e3\u3002 3 RADOS\u8be6\u7ec6\u4ecb\u7ecd 3.1 \u6269\u5c55\u673a\u7fa4 1\uff0eCluster Map \u5b58\u50a8\u673a\u7fa4\u7684\u7ba1\u7406\uff0c\u552f\u4e00\u7684\u9014\u5f84\u662fCluster Map\u901a\u8fc7\u5bf9Monitor Cluster\u64cd\u4f5c\u5b8c\u6210\u3002Cluster Map\u662f\u6574\u4e2aRADOS\u7cfb\u7edf\u7684\u6838\u5fc3\u6570\u636e\u7ed3\u6784\uff0c\u5176\u4e2d\u6307\u5b9a\u4e86\u673a\u7fa4\u4e2d\u7684OSDs\u4fe1\u606f\u548c\u6240\u6709\u6570\u636e\u7684\u5206\u5e03\u60c5\u51b5\u3002\u6240\u6709\u6d89\u53ca\u5230RADOS\u7cfb\u7edf\u7684Storage\u8282\u70b9\u548cClients\u90fd\u6709\u6700\u65b0epoch\u7684Cluster Map\u526f\u672c\u3002\u56e0\u4e3aCluster Map\u7684\u7279\u6b8a\u6027\uff0cClient\u5411\u4e0a\u63d0\u4f9b\u4e86\u975e\u5e38\u7b80\u5355\u7684\u63a5\u53e3\u5b9e\u73b0\u5c06\u6574\u4e2a\u5b58\u50a8\u673a\u7fa4\u62bd\u8c61\u4e3a\u5355\u4e00\u7684\u903b\u8f91\u5bf9\u8c61\u5b58\u50a8\u7ed3\u6784\u3002 Cluster Map\u7684\u66f4\u65b0\u7531OSD\u7684\u72b6\u6001\u53d8\u5316\u6216\u8005\u5176\u4ed6\u4e8b\u4ef6\u9020\u6210\u6570\u636e\u5c42\u7684\u53d8\u5316\u9a71\u52a8\uff0c\u6bcf\u4e00\u6b21Cluster Map\u66f4\u65b0\u90fd\u9700\u8981\u5c06map epoch\u589e\u52a0\uff0cmap epoch\u4f7fCluster Map\u5728\u6240\u6709\u8282\u70b9\u4e0a\u7684\u526f\u672c\u90fd\u4fdd\u6301\u540c\u6b65\uff0c\u540c\u65f6\uff0cmap epoch\u53ef\u4ee5\u4f7f\u4e00\u4e9b\u8fc7\u671f\u7684Cluster Map\u80fd\u591f\u901a\u8fc7\u901a\u4fe1\u5bf9\u7b49\u8282\u70b9\u53ca\u65f6\u66f4\u65b0\u3002 \u5728\u5927\u89c4\u6a21\u7684\u5206\u5e03\u5f0f\u7cfb\u7edf\u4e2d\uff0cOSDs\u7684failures\/recoveries\u662f\u5e38\u89c1\u7684\uff0c\u6240\u4ee5\uff0cCluster Map\u7684\u66f4\u65b0\u5c31\u6bd4\u8f83\u9891\u7e41\uff0c\u5982\u679c\u5c06\u6574\u4e2aCluster Map\u8fdb\u884c\u5206\u53d1\u6216\u5e7f\u64ad\u663e\u7136\u4f1a\u9020\u6210\u8d44\u6e90\u7684\u6d6a\u8d39\uff0cRADOS\u91c7\u7528\u5206\u53d1incremental map\u7684\u7b56\u7565\u907f\u514d\u8d44\u6e90\u6d6a\u8d39\uff0c\u5176\u4e2dincremental &hellip; <a href=\"https:\/\/www.strongd.net\/?p=608\" class=\"more-link\">Continue reading <span class=\"screen-reader-text\">\u5206\u5e03\u5f0f\u6587\u4ef6\u7cfb\u7edfCeph\u8c03\u78141 \u2013 RADOS<\/span><\/a><\/p>\n","protected":false},"author":2,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[1,11],"tags":[202],"class_list":["post-608","post","type-post","status-publish","format-standard","hentry","category-java","category-11","tag-202"],"_links":{"self":[{"href":"https:\/\/www.strongd.net\/index.php?rest_route=\/wp\/v2\/posts\/608","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.strongd.net\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.strongd.net\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.strongd.net\/index.php?rest_route=\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/www.strongd.net\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=608"}],"version-history":[{"count":2,"href":"https:\/\/www.strongd.net\/index.php?rest_route=\/wp\/v2\/posts\/608\/revisions"}],"predecessor-version":[{"id":610,"href":"https:\/\/www.strongd.net\/index.php?rest_route=\/wp\/v2\/posts\/608\/revisions\/610"}],"wp:attachment":[{"href":"https:\/\/www.strongd.net\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=608"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.strongd.net\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=608"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.strongd.net\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=608"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}