这是小弟第一次发文,先给个位拜个晚年
祝大伙猪年诸事大吉^^

well,言归正传,开始说说小弟遭遇的问题,看看有人可以提供意见吗?
配置：
1.两台 OS/400 (i5 520, iSeries820)要上DS6800 (1750-511)
2.DDMs：新增300GB x 8 (原Server enclosure上已经有300GB x8 DDMs)
3.RAID Type：R5 ( 6+P+S)
4.Available Capacity：1,576GB (New exrent pools)
5.DS6800 Microcode：更新至6.2.2.49 (5.2.2.133)ß此version是IBM website建议使用中的DS6800更新的version
6.Volumes：Create 141.1GB x 11 volumes(From New exrent pools)
7.VG：Creat 两个VG (VG520 , VG820),个别配置141.1 x 5 volumes(客户要求每台使用700GB)

问题：
1.Mapping给520的700GB正常使用中,可是Mapping给820的700GB,820没能认到?
(事实上,在Mapping Host时,Storage Manger中只看到520的HBA卡,820是手动key WWN的,怀疑820上的2766卡有问题,要不就是PTFs没Update完全)

2.一个Pools给两台OS/400使用,是否会有Performance上的问题?
(会这样认为是因为先让520使用了一阵子,之后才上820,可是在Mapping VG820给820时,发生520 联机Performanc变慢的现象,为免影响520运作,当下就马上将Mapping取消后,520状况就正常了,所以担心共享一个Pools,是否会彼此影响?还是在Mapping时,短暂的现象?)

3.查了redbook,每张HBA有限制的使用Capacity,不知700GB是否会有问题?
Capacity per I/O Adapter:
I/O Adapter   I/O Processor Capacity per IOA     Rule of thumb
2787          2844 1022/AD                       500GB
2766          2844 798/AD                        400GB
2766          2843 644/AD                        320GB
5760          2844 See note***
Note: ***Size the same capacity per adapter 5760 as for 2787 on 2844. For transfer sizes larger than 16KB, size about 50% more capacity than for adapter 2787.

恳请大伙把脉诊断看看:
1.820无法上DS6800的可能原因是?
2.两台OS/400共享一个Pools,会有Performance上的问题吗?
谢谢!!

注:
已check过:
1.820与DS6800的cable
2.820已上到V5R3

准备做:
1.820 PTFs与Fix的update:
因为redbook提到:
   The System i must be running V5R2, V5R3 (i5/OS) or later level of OS/400. In addition, at the time of writing, the following PTFs were required:
_ V5R2: MF33327, MF33301, MF33469, MF33302, SI14711, and SI14754
_ V5R3: MF33328, MF33845, MF33437, MF33303, SI14690, SI14755, and SI14550
   Prior to attaching the DS6000 to System i, check for the latest PTFs, which might have superseded those shown here.


因为IBM website提到:http://www-912.ibm.com/e_dir/eserverprereq.nsf/HW2/Hardware~iSeries~2766?OpenDocument

                         OS information
Code type/OS      Version/Release          Fix level            Other information
i5/OS           V5R3                 C4153530  
server firmware   FW2.1.0               SF210_028

注:问过OS/400专家,820没有server firmware

2.更新DS6800 Microcode至最新version,不过,有点担心不想做这动作咧

3.更换新HBA卡为2787(东西还没到,暂时无法做)

[ 本帖最后由 jimmyp 于 2007-3-2 15:47 编辑 ]

更新过PTFs,DS6800还是没能认到820上的HBA卡
怎么办,头痛…..
看来只有卡坏的可能了

zone正确？
switch上可以看到820卡的wwn？
不会400，不会6800，哈。

先谢谢你的回答^^

没上SAN Switch,直接接上DS6800(Host Port)
400有另外一位同仁处理,应该可以handle
只是没其它卡可测,一直觉得2766卡有问题
因为6800 host port与820的2766卡的link灯号都在闪(表示没Link)
唉,搞不起来,要回家吃自己了….呜…..

400上的Storage IOA,2766卡的Status是Operational
看来好像又正常….
注: HBA=2766 + IOP=2843

卡有什么al/pt2pt的属性吗？

嗯,这个地方到没注意到….
400上好像看不到相关的information
刚才问过400的人,表示2766没能改参数
不过,大哥提醒小弟了,可以试试上SAN Switch看看
感恩咧^^

你开始不是说直接连接6800的吗？

建议你把整个链路上的东西都check一次。
试着回答你的问题，哈。
1、可以肯定的是，如果只做mapping就会影响性能，那18m的stg团队可以洗洗睡了。不会这样的。你可以再次试试，哈。mapping只是建立一个映射关系。
看帖子，好像给6800升级过firmware，是否验证过hba卡驱动、微码与存储微码的兼容性？

2、总共增加8个hard disk，7个盘可用，6个盘参与数据I/O。性能不要有太高期望。两个server如果业务时间段分开，性能还好，如果有overlap，那可能更差。

3、hba卡与容量限制的问题，没有摸过400，不知道情况。可以查查资料看看。

4、400和6000的硬件是一致的，应该有server firmware的。

愿你早日解决问题，哈。

刚才扫了一下6000的配置指南。
第3个问题，不应该是卡有容量的限制。Rule of thumb项下的数字，应该可以理解为每块卡能承载的负载。同样，520能用700G，820为什么不可以？卡有差别？那700G也超过了2787卡的最大值。不过下面的文字，倒让我感觉真的有容量限制，晕。。。

看资料的时候发现，i上的navigator倒是挺方便的。有机会还真想摸摸i机，哈哈。

Size and number of LUNs
As discussed in “Logical volume sizes” on page 340, OS/400 can only use fixed logical
volume sizes. As a general rule of thumb, we recommend that you should configure more
logical volumes than actual DDMs. As a minimum, we recommend 2:1. For example, with 73
GB DDMs, you should use a maximum size of 35.1GB LUNs. The reason for this is that
OS/400 does not support command tag queuing. Using more, smaller LUNs can reduce I/O
queues and wait times by allowing OS/400 to support more parallel I/Os.
From the values in Table 18-2, you can calculate the number of System i Fibre Channel
adapters for your required System i disk capacity. As each I/O adapter can support a
maximum of 32 LUNs, divide the capacity per adapter by 32 to give the approximate average
size of each LUN.
For example, assume you require 2TB capacity and are using 2787 I/O adapters with 2844
I/O processors. If you know the access density, calculate the capacity per I/O adapter, or use
the rule-of-thumb. Let’s assume the rule-of-thumb of 500GB per adapter. In this case, we
would require four I/O adapters to support the workload. If we were able to have variable
LUNs sizes, we could support 32 15.6GB LUNs per I/O adapter. However, since OS/400 only
supports fixed volume sizes, we could support 28 17.5GB volumes to give us approximately
492GB per adapter.

QUOTE:

原帖由 wildhorse 于 2007-3-2 22:26 发表
建议你把整个链路上的东西都check一次。
试着回答你的问题，哈。
1、可以肯定的是，如果只做mapping就会影响性能，那18m的stg团队可以洗洗睡了。不会这样的。你可以再次试试，哈。mapping只是建立一个映射 ...

1、
“可以肯定的是，如果只做mapping就会影响性能，那18m的stg团队可以洗洗睡了。不会这样的。你可以再次试试，哈。mapping只是建立一个映射关系。”


→是阿,没道理一个Pools只给一台Server使用,除非400特别,但也没找到相关information,应该只是Create Volumes后,做Mapping时,短暂的现象,因为Mapping给820时,6800上没能认出820上的2766卡,当时是手动key WWPN试着看看可以成功否,不知这样认不到卡,又硬Mapping的动作,是否也有影响?
头痛,平常时间不给测,因为上次的动作影响到Mapping好的520,user要求假日,没人使用时,才能测…呜…


“是否验证过hba卡驱动、微码与存储微码的兼容性？”
→上IBM Web site:http://www-01.ibm.com/systems/support/storage/config/hba/index.wss
选择6800+i5/os+400/os
再选择2766,得到结果如下:
http://www-01.ibm.com/systems/support/storage/config/hba/hbaViewDetails.wss?hba_id=3
看来hba卡驱动、微码与存储微码,应该都ok

2、
“总共增加8个hard disk，7个盘可用，6个盘参与数据I/O。性能不要有太高期望。两个server如果业务时间段分开，性能还好，如果有overlap，那可能更差。”

→没得选择,6800最多只能8个hard disk(2 Sites)做1个Raid

3、
“hba卡与容量限制的问题，没有摸过400，不知道情况。可以查查资料看看。”
→这点应该是确定的,只是不知超过对performane会有多大的影响,没看到这方面的information…..
AboutCapacity per FC adapter:
Determine needed number of FC adapters by calculating how much capacity can one
adapter handle with customer’s workload. Calculate Access Density of i5/OS workload as
is described in Appendix... and use the following formulas:
–For FC adapter 2787 on IOP 2844 use: (1022 / AD) GB per FC adapter
–For FC adapter 2766 on IOP 2844 use: (798 / AD) GB per FC adapter
–For FC adapter 2766 on IOP 2843 use: (644 / AD) GB per FC adapter
If AD and blocksize are not known use save estimation AD = 2, so size:
–For FC adapter 2787 on IOP 2844 use: 511 GB per FC adapter
–For FC adapter 2766 on IOP 2844 use: 399 GB per FC adapter
–     For FC adapter 2766 on IOP 2843 use: 322 GB per FC adapter


4、
“400和6000的硬件是一致的，应该有server firmware的。”
→400专家说:520有,820没
另外,以下是找到的有关400上6800配置时要注意的地方,看到当初配置有问题,要从配置了
头痛,520已经在使用了…..唉….
Anyway,先谢谢你给的建议^^

To improve the availability of fibre-channel disks(when there is more than one 2766, 2787, or 5760 I/O adapter),use the following configuration guidelines to help you achieve good DS6000 iSeries I/O performance: Spread iSeries LUNs across as many logical ranks as possible on different DS6000 processors:
–Use no more than 500 I/O operations per second per rank.
–Assign I/O critical iSeries applications to DDMs with speeds of 15000 rpm or more.
–Create ranks of eight DDMs, if you are formatting RAID 5 ranks.v Use the Disk Magic
Analysis tool to calculate the smallest LUNs size as possible to allow more iSeries-active I/O, without effecting device response times. Assign no more than 500GB of storage to each iSeries I/O adapter. For SAN attachments, do not connect more than four iSeries I/O adapters to a single I/O attachment port. v Use only one fibre-channel IOP/IOA pair per multiadapter bridge. Do not use more than six IOP/IOA pairs per high-speed loop (HSL). Use 64-bit, 133 MHz PCI-X slots.

*Sizing production workload:
Consider the following guidelines when sizing for production workload:
n73 GB 15 K DDMs
nUse Table 5-4 for calculating the number of needed ranks. Also, for transaction workload with about 70% reads you may consider the following guidelines:
–50 IO/sec per DDM for 15 K rpm disks (not including spare and parity drives)
–35 IO/sec per DDM for 10 K rpm disks (not including spare and parity drives)
nNo more than 6 enclosures (base enclosure plus 5 expansion enclosures) for DS6800 connected to an i5 system. If customer’s requirements exceed this capacity we recommend to consider DS81000.
nCapacity per FC adapter
Determine needed number of FC adapters by calculating how much capacity can one
adapter handle with customer’s workload. Calculate Access Density of i5/OS workload as is described in Appendix... and use the following formulas:
–For FC adapter 2787 on IOP 2844 use: (1022 / AD) GB per FC adapter
–For FC adapter 2766 on IOP 2844 use: (798 / AD) GB per FC adapter
–For FC adapter 2766 on IOP 2843 use: (644 / AD) GB per FC adapter
If AD and blocksize are not known use save estimation AD = 2, so size:
–For FC adapter 2787 on IOP 2844 use: 511 GB per FC adapter
–For FC adapter 2766 on IOP 2844 use: 399 GB per FC adapter
–For FC adapter 2766 on IOP 2843 use: 322 GB per FC adapter
nDefine LUNs of size 35.1 GB or 17.54 GB.
nIf DS6000 will be shared among multiple i5/OS systems, consider to dedicate ranks to each i5/OS production workload. Consider sharing ranks only of one of the workloads is very insignificant. For example, share ranks if IO rate of light workload doesn’t exceed 10% of IO rate of important workload.
nIf DS6000 will be shared among one or more iSeries systems and one or more open systems consider the guidelines for multiple i5/OS workloads, which are listed above.
nIf DS6000 is connected to i5 system via switches size 2 iSeries FC adapters per one DS6000 FC port.
nWe recommend to use RAID ranks of 8 DDMs. Only in case 4 DDMs in a RAID rank will provide better balancing you may define 4 DDM ranks.
nDefine extent pools so that the LUNs of each iSeries system are evenly spread between the two DS6000 clusters.

[ 本帖最后由 jimmyp 于 2007-3-3 13:17 编辑 ]