[对于XR的升级，就我现在所知的有以下几种，如果还有其他的方法，到时再更新]
1. Golden Disk升级
类似于Ghost系统，在现有XR设备（version A）上直接把整个系统文件backup到disk卡中，然后把这个disk卡放到要升级的机箱上（version B），指定boot位置并重启，就可以直接用新系统了（version A）。省了安装系统的时间。但由于需要指定boot路径，另外还需要copy MBI文件到bootflash，所以操作起来感觉不是非常方便。
2013-9-18 更新:
在2010-9-24 更新中的第3和第4小点中，已经总结了一些GoldenDisk的内容。这里再提一下，如果从老版本升级到新版本，如3.6.3到4.0.4，又需要用FAT32，那么一定要把MBI拷贝到bootflash，否则无法识别FAT32；另外还要记住升级前要升级rommona/b的版本，原因下面有解释。
在3.6.3上升级rommon的步骤如下：
1. 从官网下载较新的rommon tar文件
2. 拷贝到disk0中
3. 执行upgrade rommon a/b all disk0
如果客户安装了很多SMU，又不知道正确的MBI路径，那么可以在老系统中把默认的MBI路径拷贝到bootflash中，然后启动，设备会自动boot这个MBI，如果MBI不正确，它会自动把正确的MBI文件copy到bootflash中，然后重启并用新MBI启动，如下所示：
Install (Node Preparation): Booting with committed software
Install (Node Preparation): Checking free space on /bootflash/
Install (Node Preparation): Copying MBI from /disk0/hfr-os-mbi-4.0.4.CSCuc17728-1.0.0/mbihfr-rp.vm
Install (Node Preparation): Copying MBI to /bootflash/disk0/hfr-os-mbi-4.0.4.CSCuc17728-1.0.0/mbihfr-rp.vm
insthelper: This node was booted with an incorrect MBI (/bootflash/disk0/hfr-os-mbi-4.0.4/mbihfr-rp.vm), rebooting self with correct boot image (/bootflash/disk0/hfr-os-mbi-4.0.4.CSCuc17728-1.0.0/mbihfr-rp.vm).
2. Turboboot 升级
完全重新安装，在Rommon模式中download mini PIE文件，感觉像之前IOS环境下的启动新版本那样。只不过XR是要安装，而IOS直接调用。
3. 直接在原有系统上升级
打个比方说，在windows xp的基础上直接升级到windows 7，这样即使windows 7装完了，在新系统中仍然有老系统的东西。所以一般不推荐用这种方法。对于disk存储介质，最好只有1种操作系统，这样稳定些。
2013-11-29 更新：
在现有系统升级又分两种方式：
一种是直接安装PIE文件，如果你的版本是A，要升级B，那么直接active B的PIE即可，但要注意，A和B的PIE要保持一致才能安装，否则会有兼容，如下所示(注：现有系统升级后，配置是保留的)：
RP/0/RP0/CPU0:LAB-CRS-1(admin)#show ins ac sum
Wed Nov 20 08:56:25.113 UTC
Default Profile:
Admin Resources
SDRs:
Owner
Active Packages:
disk0:hfr-mpls-px-4.0.4
disk0:hfr-mgbl-px-4.0.4
disk0:hfr-mini-px-4.0.4
RP/0/RP0/CPU0:LAB-CRS-1(admin)#install activate disk0:hfr-mini-px-4.2.4 sync
Wed Nov 20 08:56:33.503 UTC
Install operation 34 '(admin) install activate disk0:hfr-mini-px-4.2.4 synchronous' started by user 'cisco' via CLI at 08:56:33 UTC Wed Nov 20 2013.
Error: Cannot proceed with the activation because of the following package incompatibilities:
Error: hfr-mgbl-supp-4.0.4 needs iosxr-fwding-4.0.4, or equivalent, to be active on the same nodes.
Error: hfr-mgbl-supp-4.0.4 needs hfr-fwding-4.0.4, or equivalent, to be active on the same nodes.
Error: hfr-mgbl-supp-4.0.4 needs hfr-base-4.0.4, or equivalent, to be active on the same nodes.
Error: hfr-mgbl-supp-4.0.4 needs iosxr-infra-4.0.4, or equivalent, to be active on the same nodes.
Error: hfr-mgbl-supp-4.0.4 needs hfr-os-support-4.0.4, or equivalent, to be active on the same nodes.
Error: hfr-mgbl-supp-4.0.4 needs iosxr-os-4.0.4, or equivalent, to be active on the same nodes.
Error: iosxr-mgbl-4.0.4 needs iosxr-routing-4.0.4, or equivalent, to be active on the same nodes.
Error: iosxr-mgbl-4.0.4 needs iosxr-fwding-4.0.4, or equivalent, to be active on the same nodes.
Error: iosxr-mgbl-4.0.4 needs iosxr-infra-4.0.4, or equivalent, to be active on the same nodes.
Error: iosxr-mgbl-4.0.4 needs iosxr-os-4.0.4, or equivalent, to be active on the same nodes.
Error: iosxr-mpls-4.0.4 needs iosxr-routing-4.0.4, or equivalent, to be active on the same nodes.
Error: iosxr-mpls-4.0.4 needs iosxr-fwding-4.0.4, or equivalent, to be active on the same nodes.
Error: iosxr-mpls-4.0.4 needs iosxr-infra-4.0.4, or equivalent, to be active on the same nodes.
Error: iosxr-mpls-4.0.4 needs iosxr-os-4.0.4, or equivalent, to be active on the same nodes.
Error: Suggested steps to resolve this:
Error: - check the installation instructions.
Error: - activate or deactivate the specified packages on the specified nodes.
Install operation 34 failed at 08:56:40 UTC Wed Nov 20 2013.
RP/0/RP0/CPU0:LAB-CRS-1(admin)#
RP/0/RP0/CPU0:LAB-CRS-1(admin)#install add tftp://10.75.41.57/hfr-mgbl-px.pie-4.2.4 sync
Wed Nov 20 08:57:39.796 UTC
Install operation 35 '(admin) install add /tftp://10.75.41.57/hfr-mgbl-px.pie-4.2.4 synchronous' started by user 'cisco' via CLI at 08:57:39 UTC Wed Nov 20 2013.
Info: The following package is now available to be activated:
Info:
Info: disk0:hfr-mgbl-px-4.2.4
Info:
Info: The package can be activated across the entire router.
Info:
Install operation 35 completed successfully at 08:58:02 UTC Wed Nov 20 2013.
RP/0/RP0/CPU0:LAB-CRS-1(admin)#install activate disk0:hfr-mini-px-4.2.4 disk0:hfr-mgbl-px-4.2.4 sync
Wed Nov 20 08:58:46.818 UTC
Install operation 36 '(admin) install activate disk0:hfr-mini-px-4.2.4 disk0:hfr-mgbl-px-4.2.4 synchronous' started by user 'cisco' via CLI at 08:58:46 UTC Wed Nov 20
2013.
Error: Cannot proceed with the activation because of the following package incompatibilities:
Error: iosxr-mpls-4.0.4 needs iosxr-routing-4.0.4, or equivalent, to be active on the same nodes.
Error: iosxr-mpls-4.0.4 needs iosxr-fwding-4.0.4, or equivalent, to be active on the same nodes.
Error: iosxr-mpls-4.0.4 needs iosxr-infra-4.0.4, or equivalent, to be active on the same nodes.
Error: iosxr-mpls-4.0.4 needs iosxr-os-4.0.4, or equivalent, to be active on the same nodes.
Error: Suggested steps to resolve this:
Error: - check the installation instructions.
Error: - activate or deactivate the specified packages on the specified nodes.
Install operation 36 failed at 08:58:54 UTC Wed Nov 20 2013.
RP/0/RP0/CPU0:LAB-CRS-1(admin)#
RP/0/RP0/CPU0:LAB-CRS-1(admin)#install deactivate disk0:hfr-mpls-px-4.0.4
Wed Nov 20 08:59:20.777 UTC
Install operation 37 '(admin) install deactivate disk0:hfr-mpls-px-4.0.4' started by user 'cisco' via CLI at 08:59:20 UTC Wed Nov 20 2013.
Info: Install Method: Parallel Process Restart
The install operation will continue asynchronously.
RP/0/RP0/CPU0:LAB-CRS-1(admin)#show ins ac sum
Wed Nov 20 09:00:36.442 UTC
Default Profile:
Admin Resources
SDRs:
Owner
Active Packages:
disk0:hfr-mgbl-px-4.0.4
disk0:hfr-mini-px-4.0.4
RP/0/RP0/CPU0:LAB-CRS-1(admin)#
RP/0/RP0/CPU0:LAB-CRS-1(admin)#install activate disk0:hfr-mini-px-4.2.4 disk0:hfr-mgbl-px-4.2.4 sync
Wed Nov 20 09:00:50.380 UTC
Install operation 38 '(admin) install activate disk0:hfr-mini-px-4.2.4 disk0:hfr-mgbl-px-4.2.4 synchronous' started by user 'cisco' via CLI at 09:00:50 UTC Wed Nov 20
2013.
Info: This operation will reload the following nodes in parallel:
Info: 0/7/CPU0 (LC) (SDR: Owner)
Info: 0/15/SP (MSC-DRP-SP) (Admin Resource)
Info: 0/15/CPU0 (LC) (SDR: Owner)
Info: 0/RP0/CPU0 (CRS16-RP-x86) (SDR: Owner)
Info: 0/FC0/SP (SP) (Admin Resource)
Info: 0/FC1/SP (SP) (Admin Resource)
Info: 0/AM0/SP (SP) (Admin Resource)
Info: 0/AM1/SP (SP) (Admin Resource)
Info: 0/SM2/SP (140G-Fabric-SP-B) (Admin Resource)
Info: 0/SM3/SP (140G-Fabric-SP-B) (Admin Resource)
Info: 0/SM4/SP (140G-Fabric-SP-B) (Admin Resource)
Info: 0/SM5/SP (140G-Fabric-SP-B) (Admin Resource)
Info: 0/SM6/SP (140G-Fabric-SP-B) (Admin Resource)
Info: 0/SM7/SP (140G-Fabric-SP-B) (Admin Resource)
Proceed with this install operation (y/n)? [y]
Info: Install Method: Parallel Reload
Info: The changes made to software configurations will not be persistent across system reloads. Use the command '(admin) install commit' to make changes persistent.
Info: Please verify that the system is consistent following the software change using the following commands:
Info: show system verify
Info: install verify packages
Install operation 38 completed successfully at 09:06:24 UTC Wed Nov 20 2013.
RP/0/RP0/CPU0:LAB-CRS-1(admin)#
另一种方式是直接安装下载后的TAR文件，或者自己做的一个TAR文件，以后有机会抓着部分信息，在更新上去。
关于rommon的一些问题
CRS一共有两个rommon，rommon A和rommon B。平时启动都是rommon B，rommon A只做一些工厂检测用。[如果rommon A的版本不低于1.52（这是CRS的需求，9k的另需要确认），就不用升级rommon A。因为每次启动都会由rommon A 去disk卡中check一些信息，感觉像是rommon在加载一些命令集（monlib ?），使得如果rommonB不能启动，或是坏了，可以直接从rommon A启动进行检测。如果版本低于1.52，rommon A就会不识别fat32的disk，所以就停在那里无法继续进行加载rommon B，如果是fat16，就没有任何问题]
完整阅读
