为何无法在AIX下创建大小超过100GB的文件系统

kennyhuang

各位大虾：\r\n                今日小弟在一项目实施中偶遇一个问题，请教各位大虾赐教.\r\n            系统环境：     主机：           P610 6C1（1*SSA卡4-P ）\r\n                                     操作系统：   AIX 4.3.3.09\r\n                                     存储：          7133磁盘阵列（4*36.4硬盘）\r\n            在操作系统下，将7133的4块36.4GB的硬盘做成1个RAID-5，HDISK1大小为109.3GB，然后将HDISK1全部作成DATAVG（PPSIZE为128MB），小弟想在DATAVG上作一个大小为100GB的大文件系统，用于装载DATABASE的数据文件。\r\n          但是当我使用SMITT创建文件系统（JFS）时，SMITT报错，显示该文件系统的FRAGMENT SIZE为4096bytes，PP不能超过512个。小弟不解，请问：\r\n1、如何创建超过100GBSIZE的大文件系统？\r\n2、FRAGMENT参数代表什么含义，它与文件系统的大小有什么联系？\r\n\r\n

Fenng

定制你的核心吧\r\n你的核心限制你当前的文件系统

kennyhuang

Mr Fenng:\r\n               你好！非常感谢你的回复，但是小弟对AIX实在不是很熟，能否再告之的详细一点？如何实现你所说的定制核心？？

Fenng

查看核心参数lsattr -E -l sys0\r\n设置核心：chdev\r\n\r\n我的经验也没有，很抱歉\r\n为什么一定要那么大的文件系统？？\r\n所有的鸡蛋放到一个篮子里，会有问题的

kennyhuang

你能肯定是核心参数的问题吗？？？

slls

LV有一个属性：MAXIMUM NUMBER of LOGICAL PARTITIONS缺省是512\r\n你其中报的那个错PP不能超过512个应该就是指的它。\r\n建议你先建一个lv并把它设大些，然后再在其之上建文件系统。\r\n根据IBM红皮书system administrator上说得：\r\n文件系统的最大值是:\r\nFRAGMENT SIZE * 2的28次方，应该能满足你的需要

kennyhuang

先建LV的方法我也用过，我先建了一个LV，它的MAX PPs为512个，我在其上无法创建大于512个PP的FS，后来，我在此LV上将MAX PPs修改为了800个，并且该LV的大小扩展到了600个PP（每个PP的大小为128MB），但是原来的FS仍然无法超过512个PP的限制

slls

那就试试重新建一个VG， Physical partition SIZE in megabytes 选256，\r\n另外，我建的文件系统LV的LP远远超过512了，难道是这个原因？

kennyhuang

我主要想弄明白，这个问题跟创建FS时的哪些参数有关系，或者跟LV、VG的参数有关系？以及具体关系（算法）是什么，我想这才是把这个问题弄明白的意义所在。

slls

《System Management Concepts: Operating System and Devices 》[/COLOR] \r\n\r\nUnderstanding Journaled File System Size Limitations\r\nThe maximum size for a journaled file system (JFS) is defined when the file system is created. When you create a JFS, there are five significant issues to consider:\r\n\r\nNumber of i-nodes \r\nAllocation group size \r\nFile system fragment addressability \r\nJournaled file system log size \r\nMaximum journaled file system size\r\nNumber of Inodes \r\nThe total number of i-nodes in a file system limits the total number of files and the total size of the file system. The JFS provides the nbpi (number of bytes per i-node) parameter that affects the number of i-nodes in a file system. JFS supports nbpi values of 512, 1024, 2048, 4096, 8192, 16384, 32768, 65536, and 131072. The values 32768, 65536, and 131072 only apply to AIX Version 4.2 or later.\r\n\r\nOne i-node is created for every NBPI bytes of allocation group space allocated to the filesystem. An allocation group can be partially allocated, though the full number of i-nodes per allocation group is still allocated. NBPI is inversely proportional to the total number of i-nodes in a file system.\r\n\r\nThe JFS restricts all file systems to 16M (224) i-nodes.\r\n\r\nAllocation Group Size\r\nAIX Version 4.2 or later supports various allocation group sizes. The JFS segregates file system space into groupings of i-nodes and disk blocks for user data. These groupings are called allocation groups. The allocation group size can be specified when the file system is created. The allocation group sizes are 8M, 16M, 32M, and 64M. Each allocation group size has an associated npbi range. The ranges are defined by the following table:\r\n\r\nAllocation Group\r\nSize in Megabytes      Allowable NBPI Values\r\n 8                     512, 1024, 2048, 4096, 8192, 16384 \r\n16                          1024, 2048, 4096, 8192, 16384, 32768 \r\n32                                2048, 4096, 8192, 16384, 32768, 65536 \r\n64                                      4096, 8192, 16384, 32768, 65536, 131072 \r\nFile System Fragment Addressability \r\nThe JFS supports four fragment sizes of 512, 1024, 2048, and 4096 byte units of contiguous disk space. The JFS maintains fragment addresses in i-nodes and indirect blocks as 28-bit numbers. Each fragment must be addressable by a number from 0 to (22.\r\n\r\nJournaled File System Log Size Issues \r\nAnother size-related issue is the size of the JFS log. In most instances, multiple journaled file systems use a common log configured to be 4MB in size. For example, after initial installation, all file systems within the root volume group use logical volume hd8 as a common JFS log. The default logical volume partition size is 4MB, and the default log size is one partition, therefore, the root volume group normally contains a 4MB JFS log. When file systems exceed 2GB or when the total amount of file system space using a single log exceeds 2GB, the default log size may not be sufficient. In either case, the log sizes should be scaled upward as the file system size increases. The JFS log is limited to a maximum size of 256MB.\r\n\r\nMaximum Journaled File System Size \r\nThe maximum JFS size is defined when the file system is created. For example, selecting an nbpi ratio of 512 will limit the file system to a size of 8GB (512 * 224 = 8GB). When creating a JFS file system, the factors listed above (nbpi, fragment size, and allocation group size) need to be weighed carefully. The filesystem size limitation is the minimum of NPBI * 2^24 or Fragment Size * 2^28[/COLOR]\r\n你一定是NPBI选小了！