EAS部署：linux 下安裝EAS後啓動不了服務

一、問題

同事將原Windows上的EAS服務器移植到CentOS後，在打開管理控制檯--啓動服務時卡住了，等了20分鐘服務都沒啓動了。

查看了下admin/logs/admin.log，主要內容以下：

 1 Find java home: /kingdee/eas/ibm-jdk/jre
 2 Set java home: /kingdee/eas/ibm-jdk
 3 [2017-12-07 13:35:14,887 INFO]Execute base path [/kingdee/eas/ibm-jdk/bin] cmd [java, -Xmx4096m, -Xms2048m, -XX:MaxPermSize=768m, -XX:PermSize=768m, -version] to check enough JVMHeap.
 4 [2017-12-07 13:35:15,122 INFO]Process exitValue: 0
 5 [2017-12-07 13:35:15,123 INFO]linuxOSTmpPath is [/kingdee/eas/admin/workspace]
 6 [2017-12-07 13:35:15,126 INFO]Begin execute command [df -k /kingdee/eas/admin/workspace | awk '{FS="[ \t]+"}{print $4;}' |  sed '1d'] !
 7 [2017-12-07 13:35:15,160 INFO]Execute command [df -k /kingdee/eas/admin/workspace | awk '{FS="[ \t]+"}{print $4;}' |  sed '1d'] success !
 8 [2017-12-07 13:35:15,161 INFO]TemporaryDirectory is [/kingdee/eas/admin/workspace] and freeTempSpace is [65463568KB].
 9 [2017-12-07 13:35:15,162 INFO]Module [apusic4.0.3] isn't exists, try find module [apusic4]
10 [2017-12-07 13:35:15,163 INFO]Module [apusic4] isn't exists, try find module [apusic]
11 [2017-12-07 13:35:15,164 INFO]Get appengine[id:1872326553] by key[apusic:127.0.0.1/6888:admin:server1]
12 [2017-12-07 13:35:15,165 INFO][/kingdee/apusic] versio is not 403
13 [2017-12-07 13:35:15,168 INFO]Module [apusic4.0.3] isn't exists, try find module [apusic4]
14 [2017-12-07 13:35:15,168 INFO]Module [apusic4] isn't exists, try find module [apusic]
15 [2017-12-07 13:35:15,169 INFO]Get appengine[id:1872326553] by key[apusic:127.0.0.1/6888:admin:server1]
16 [2017-12-07 13:35:15,170 INFO][/kingdee/apusic] versio is not 403
17 [2017-12-07 13:35:15,172 INFO]Module [apusic4.0.3] isn't exists, try find module [apusic4]
18 [2017-12-07 13:35:15,172 INFO]Module [apusic4] isn't exists, try find module [apusic]
19 [2017-12-07 13:35:15,173 INFO]Get appengine[id:1872326553] by key[apusic:127.0.0.1/6888:admin:server1]
20 [2017-12-07 13:35:15,178 INFO]Start application server [/kingdee/eas/server/profiles/server1/bin] cmd[/bin/sh, -C, startserver.sh, nohup].....
21 [2017-12-07 13:35:15,216 ERROR]rm: cannot remove `/kingdee/apusic/common/jsf-api.jar': No such file or directory
22 [2017-12-07 13:35:15,216 INFO]isServer1 : server1
23 [2017-12-07 13:35:15,227 ERROR]rm: cannot remove `/kingdee/apusic/lib/operamasks-impl.jar': No such file or directory
24 [2017-12-07 13:35:15,258 INFO]Begin to make instance template dir ./workspace....
25 [2017-12-07 13:35:15,259 INFO]../workspace exists,remove then create it...
26 [2017-12-07 13:35:15,258 ERROR]rm: cannot remove `/kingdee/apusic/lib/ext/operamasks-third-party.jar': No such file or directory
27 [2017-12-07 13:35:15,259 INFO]End to make instance template dir ./workspace....
28 [2017-12-07 13:35:15,383 ERROR]JVMJ9GC032E System configuration does not support option '-Xlp'
29 [2017-12-07 13:35:15,383 ERROR]JVMJ9VM015W Initialization error for library j9gc23(2): Failed to initialize; unable to parse command line
30 [2017-12-07 13:35:15,385 ERROR]Could not create the Java virtual machine.

從日誌中發現一個關鍵信息：JVMJ9GC032E System configuration does not support option '-Xlp'

第一反應是JVM參數配置有問題，找到 eas\server\profiles\server1\bin\set-server-env.sh配置文件中 確實有一項 配置JVM_CUSTOM_PARAMS中指定了-Xlp配置，將此配置取消後成功啓動。

 

二、分析

由於使用的爲IBM JDK，在官網找到了JVM相關參數的說明（http://lampwww.epfl.ch/java/jre-ibm-1.5/docs/zh_Hans/sdkandruntimeguide.lnx.zh_Hans.htm）：

-Xlp

    Requests the JVM to allocate the Java heap with large pages. If large pages are not available, the JVM will not start, displaying the error message GC: system configuration does not support option --> '-Xlp'. The JVM uses shmget() to allocate large pages for the heap. Large pages are supported by systems running Linux kernels v2.6 or higher, or earlier kernels where large page support has been backported by the distribution. By default, large pages are not used. See Configuring large page memory allocation.

You can enable large page support, on systems that support it, by starting |Java with the -Xlp option.
|

Large page usage is primarily intended to provide performance improvements |to applications that allocate a lot of memory and frequently access that memory. |The large page performance improvements are mainly caused by the reduced number |of misses in the Translation Lookaside Buffer (TLB). The TLB maps a larger |virtual memory range and thus causes this improvement.
|

Large page support must be available in the kernel, and enabled, |to allow Java to use large pages.
|

To configure large page memory allocation, first ensure that |the running kernel supports large pages. Check that the file /proc/meminfo |contains the following lines:
|

HugePages_Total:     <number of pages>
|HugePages_Free:      <number of pages>
|Hugepagesize:        <page size, in kB>

The number of pages available |and their sizes vary between distributions.
|

If large page support is not available in your kernel, these |lines will not exist in the /proc/meminfo file. In this case, you must install |a new kernel containing support for large pages.
|

If large page support is available, but not enabled, HugePages_Total will be 0. In this case, your administrator must enable |large page support. Check your operating system manual for more instructions.
|

For the JVM to use large pages, your system must have an adequate number |of contiguous large pages available. If large pages cannot be allocated, even |when enough pages are available, possibly the large pages are not contiguous. Configuring the number of large pages at bootup will create |them contiguously.
|

Large page allocations will only succeed if the JVM has root |access. To use large pages, either run Java as root or set the suid bit of |the Java executable.

經過以上的內容的描述，才得知是操做系統沒啓用 large page support功能才致使eas 服務啓動不了，查看了系統的/proc/meminfo，確實和文中描述一致。