首页 > 代码库 > openstack内存占用清理

openstack内存占用清理


公司使用openstack定制开发了云桌面平台提供给客户使用,架构是3台控制节点+N台计算节点+N台CEPH+cinder。使用一直较为稳定,今天部份客户反映双击云主机图标无法进入云桌面、或者需要多次点击方可进入云桌面。经过检查是因为2台控制节点内存占用过高导致,需要清理。下面是详细操作过程。

1、查看控制节点内存占用

[root@node-6 ~]# top
top - 15:29:40 up 18 days, 17:15,  1 user,  load average: 1.27, 1.83, 2.09
Tasks: 1062 total,   6 running, 1056 sleeping,   0 stopped,   0 zombie
Cpu(s): 12.2%us,  1.1%sy,  0.0%ni, 86.4%id,  0.1%wa,  0.0%hi,  0.2%si,  0.0%st
Mem:  99032136k total, 97932848k used,  1099288k free,   233100k buffers
Swap: 33554428k total,   295916k used, 33258512k free, 38017888k cached
  PID USER      PR  NI  VIRT  RES  SHR S %CPU %MEM    TIME+  COMMAND                        2926 rabbitmq  20   0 2904m 449m 2676 S 97.4  0.5   0:12.85 beam.smp                     1868 root      10 -10 44932  11m 7240 S 15.0  0.0   2149:53 ovs-vswitchd                 7982 mysql     -2   0 24.8g 2.8g 138m S  9.4  3.0   4617:11 mysqld                        ......

"Mem:  99032136k total, 97932848k used",可以看到 控制节点内存已用满


控制节点查看nova服务状态:

[root@node-6 ~]# nova-manage service list
Binary           Host                                 Zone             Status     State Updated_At
nova-consoleauth node-6.domain.tld                    internal         enabled    :-)   2017-05-04 07:26:39
nova-scheduler   node-6.domain.tld                    internal         enabled    :-)   2017-05-04 07:26:40
nova-conductor   node-6.domain.tld                    internal         enabled    :-)   2017-05-04 07:26:49
nova-cert        node-6.domain.tld                    internal         enabled    :-)   2017-05-04 07:26:47
nova-consoleauth node-7.domain.tld                    internal         enabled    :-)   2017-05-04 07:26:43
nova-scheduler   node-7.domain.tld                    internal         enabled    :-)   2017-05-04 07:26:48
nova-conductor   node-7.domain.tld                    internal         enabled    :-)   2017-05-04 07:26:48
nova-consoleauth node-8.domain.tld                    internal         enabled    :-)   2017-05-04 07:26:39
nova-scheduler   node-8.domain.tld                    internal         enabled    :-)   2017-05-04 07:26:42
nova-conductor   node-8.domain.tld                    internal         enabled    :-)   2017-05-04 07:26:48
nova-cert        node-7.domain.tld                    internal         enabled    :-)   2017-05-04 07:26:41
nova-cert        node-8.domain.tld                    internal         enabled    :-)   2017-05-04 07:26:39
nova-compute     node-4.domain.tld                    nova             disabled   :-)   2017-05-04 07:26:40
nova-compute     node-10.domain.tld                   nova             enabled    :-)   2017-05-04 07:26:47
nova-compute     node-15.domain.tld                   nova             enabled    :-)   2017-05-04 07:26:44
nova-compute     node-16.domain.tld                   nova             enabled    :-)   2017-05-04 07:26:45
nova-compute     node-11.domain.tld                   nova             enabled    :-)   2017-05-04 07:26:41
nova-compute     node-17.domain.tld                   nova             enabled    :-)   2017-05-04 07:26:39
nova-compute     node-5.domain.tld                    nova             enabled    :-)   2017-05-04 07:26:46
nova-compute     node-9.domain.tld                    nova             enabled    :-)   2017-05-04 07:26:41
nova-compute     node-13.domain.tld                   nova             enabled    :-)   2017-05-04 07:26:42
nova-compute     node-14.domain.tld                   nova             enabled    :-)   2017-05-04 07:26:46
nova-compute     node-12.domain.tld                   nova             enabled    :-)   2017-05-04 07:26:41
nova-compute     node-26.domain.tld                   nova             enabled    :-)   2017-05-04 07:26:48


在控制节点查看了内存占用及nova服务状态控制。node-4 libvirtd服务disabled,一般在该计算节点用“/etc/init.d/libvirtd restart”重启即可。虽然linux有内存管理机制,但目前服务器内存用满不会自动下降,却容易导致各节点通讯失败。导致用户“双击云主机图标无法进入云桌面”,只能人工介入处理清理内存。


当前的处理过程是  清理控制节点内存-->重启控制节点rabbitmy-server服务--> 重启各计算节点nova-compute服务。当然清理内存也有风险,可能导致在用用户云主机断开影响操作体验。


2、第一台控制节点清理内存,重启rabbitmy-server服务

[root@node-6 ~]# sync
[root@node-6 ~]# echo 3 > /proc/sys/vm/drop_caches
[root@node-6 ~]# top
top - 15:30:47 up 18 days, 17:16,  1 user,  load average: 1.29, 1.73, 2.04
Tasks: 1063 total,   4 running, 1059 sleeping,   0 stopped,   0 zombie
Cpu(s): 12.2%us,  1.1%sy,  0.0%ni, 86.4%id,  0.1%wa,  0.0%hi,  0.2%si,  0.0%st
Mem:  99032136k total, 59740848k used, 39291288k free,     9576k buffers
Swap: 33554428k total,   295916k used, 33258512k free,   214468k cached
  PID USER      PR  NI  VIRT  RES  SHR S %CPU %MEM    TIME+  COMMAND                         2926 rabbitmq  20   0 3222m 735m 2676 S 35.4  0.8   0:19.17 beam.smp                     
 8392 keystone  20   0  423m  66m 3524 S 20.5  0.1  15:43.19 keystone-all             
 1868 root      10 -10 44932  11m 7240 S 14.9  0.0   2150:02 ovs-vswitchd             
 7982 mysql     -2   0 24.8g 2.8g 138m S  9.3  3.0   4617:44 mysqld                           ......                                                                                                                    
 [root@node-6 ~]# /etc/init.d/rabbitmq-server restart
Restarting rabbitmq-server: FAILED - check /var/log/rabbitmq/shutdown_log, _err
RabbitMQ is going to make 3 attempts to find master node and start.
3 attempts left to start RabbitMQ Server before consider start failed.
SUCCESS
Setting policy "ha-all" for pattern "." to "{\"ha-mode\":\"all\", \"ha-sync-mode\":\"automatic\"}" with priority "0" ...
...done.
rabbitmq-server.
 
 [root@node-6 ~]# ssh node-7
Warning: Permanently added ‘node-7,192.168.10.6‘ (RSA) to the list of known hosts.
Last login: Tue May  2 09:40:05 2017 from 192.168.10.4


3、第二台控制节点清理内存:

[root@node-7 ~]# top
top - 15:30:53 up 44 days, 6 min,  1 user,  load average: 1.33, 1.44, 1.63
Tasks: 1081 total,   5 running, 1076 sleeping,   0 stopped,   0 zombie
Cpu(s): 11.3%us,  1.0%sy,  0.0%ni, 87.5%id,  0.1%wa,  0.0%hi,  0.1%si,  0.0%st
Mem:  99032136k total, 98589812k used,   442324k free,   273588k buffers
Swap: 33554428k total,   141844k used, 33412584k free, 35998936k cached
  PID USER      PR  NI  VIRT  RES  SHR S %CPU %MEM    TIME+  COMMAND                       
12728 neutron   20   0  549m 216m 2016 S 37.7  0.2  15:41.11 neutron-server              26690 keystone  20   0  327m  66m 3428 R 24.5  0.1  13:40.80 keystone-all            
26698 keystone  20   0  327m  67m 3428 S 16.9  0.1  12:57.10 keystone-all         
19852 nova      20   0  555m 197m 3332 S 15.1  0.2 502:36.04 nova-api                    
1936 root      10 -10 47608  14m 7220 S  9.4  0.0   3790:59 ovs-vswitchd                                                                                                          [root@node-7 ~]# sync  
[root@node-7 ~]# echo 3 > /proc/sys/vm/drop_caches
[root@node-7 ~]# /etc/init.d/rabbitmq-server restart
Restarting rabbitmq-server: rmdir: failed to remove `/var/run/rabbitmq‘: Directory not empty
RabbitMQ is going to make 3 attempts to find master node and start.
3 attempts left to start RabbitMQ Server before consider start failed.
SUCCESS
Setting policy "ha-all" for pattern "." to "{\"ha-mode\":\"all\", \"ha-sync-mode\":\"automatic\"}" with priority "0" ...
...done.
rabbitmq-server.


4、重启计算节点的nova-compute服务。

[root@node-7 ~]# ssh node-4
Warning: Permanently added ‘node-4,192.168.10.2‘ (RSA) to the list of known hosts.
Last login: Thu May  4 15:31:29 2017 from 192.168.10.6
[root@node-4 ~]# /etc/init.d/libvirtd restart
Stopping libvirtd daemon:                                  [FAILED]
Starting libvirtd daemon:                                  [  OK  ]
[root@node-4 ~]# /etc/init.d/openstack-nova-compute restart
Stopping openstack-nova-compute:                           [FAILED]
Starting openstack-nova-compute:                           [  OK  ]
[root@node-4 ~]# exit
logout
Connection to node-4 closed.


5、在控制节点上查看各节点nova服务状态

[root@
Binary           Host                                 Zone             Status     State Updated_At
nova-consoleauth node-6.domain.tld                    internal         enabled    :-)   2017-05-04 07:32:09
nova-scheduler   node-6.domain.tld                    internal         enabled    :-)   2017-05-04 07:32:10
nova-conductor   node-6.domain.tld                    internal         enabled    :-)   2017-05-04 07:32:16
nova-cert        node-6.domain.tld                    internal         enabled    :-)   2017-05-04 07:32:07
nova-consoleauth node-7.domain.tld                    internal         enabled    :-)   2017-05-04 07:32:13
nova-scheduler   node-7.domain.tld                    internal         enabled    :-)   2017-05-04 07:32:08
nova-conductor   node-7.domain.tld                    internal         enabled    :-)   2017-05-04 07:32:16
nova-consoleauth node-8.domain.tld                    internal         enabled    :-)   2017-05-04 07:32:09
nova-scheduler   node-8.domain.tld                    internal         enabled    :-)   2017-05-04 07:32:12
nova-conductor   node-8.domain.tld                    internal         enabled    :-)   2017-05-04 07:32:16
nova-cert        node-7.domain.tld                    internal         enabled    :-)   2017-05-04 07:32:11
nova-cert        node-8.domain.tld                    internal         enabled    :-)   2017-05-04 07:32:09
nova-compute     node-4.domain.tld                    nova             enabled    :-)   2017-05-04 07:32:05
nova-compute     node-10.domain.tld                   nova             enabled    :-)   2017-05-04 07:32:07
nova-compute     node-15.domain.tld                   nova             enabled    :-)   2017-05-04 07:32:06
nova-compute     node-16.domain.tld                   nova             enabled    :-)   2017-05-04 07:32:10
nova-compute     node-11.domain.tld                   nova             enabled    :-)   2017-05-04 07:32:12
nova-compute     node-17.domain.tld                   nova             enabled    :-)   2017-05-04 07:32:14
nova-compute     node-5.domain.tld                    nova             enabled    :-)   2017-05-04 07:32:16
nova-compute     node-9.domain.tld                    nova             enabled    :-)   2017-05-04 07:32:11
nova-compute     node-13.domain.tld                   nova             enabled    :-)   2017-05-04 07:32:12
nova-compute     node-14.domain.tld                   nova             enabled    :-)   2017-05-04 07:32:13
nova-compute     node-12.domain.tld                   nova             enabled    :-)   2017-05-04 07:32:15
nova-compute     node-26.domain.tld                   nova             enabled    :-)   2017-05-04 07:32:11

6、各节点nova服务已正常。一般来说,重启了控制的rabbitmq-server,也可以重启各计算的nova-compute服务,以重新建立连接(根据实际情况判断是否需要)。


计算节点node-5重启libvirtd及openstack-nova-compute服务

[root@node-7 ~]# ssh node-5
Warning: Permanently added ‘node-5,192.168.10.3‘ (RSA) to the list of known hosts.
Last login: Sat Apr 22 10:37:43 2017 from 192.168.10.4
[root@node-5 ~]# /etc/init.d/libvirtd restart
Stopping libvirtd daemon:                                  [  OK  ]
Starting libvirtd daemon:                                  [  OK  ]
[root@node-5 ~]# /etc/init.d/open
openstack-nova-compute  openvswitch             
[root@node-5 ~]# /etc/init.d/openstack-nova-compute restart
Stopping openstack-nova-compute:                           [  OK  ]
Starting openstack-nova-compute:                           [  OK  ]

计算节点node-9重启libvirtd及openstack-nova-compute服务

[root@node-9 ~]# /etc/init.d/libvirtd restart
Stopping libvirtd daemon:                                  [  OK  ]
Starting libvirtd daemon:                                  [  OK  ]
[root@node-9 ~]# /etc/init.d/openstack-nova-compute restart
Stopping openstack-nova-compute:                           [  OK  ]
Starting openstack-nova-compute:                           [  OK  ]

计算节点node-10重启libvirtd及openstack-nova-compute服务

[root@node-9 ~]# ssh node-10
Warning: Permanently added ‘node-10,192.168.10.9‘ (RSA) to the list of known hosts.
Last login: Tue Apr 25 16:30:26 2017 from 192.168.10.6
[root@node-10 ~]# /etc/init.d/libvirtd restart
Stopping libvirtd daemon:                                  [  OK  ]
Starting libvirtd daemon:                                  [  OK  ]
[root@node-10 ~]# /etc/init.d/openstack-nova-compute restart
Stopping openstack-nova-compute:                           [  OK  ]
Starting openstack-nova-compute:                           [  OK  ]

......重复操作。


本文出自 “福州恒达电脑” 博客,请务必保留此出处http://fzhddn.blog.51cto.com/12650899/1922090

openstack内存占用清理