1. 需求说明
glance作为openstack中p_w_picpath服务，支持多种适配器，支持将p_w_picpath存放到本地文件系统，http服务器，ceph分布式文件系统，glusterfs和sleepdog等开源的分布式文件系统上，本文，通过将讲述glance如何和ceph结合。
目前glance采用的是本地filesystem的方式存储，存放在默认的路径/var/lib/glance/p_w_picpaths下，当把本地的文件系统修改为分布式的文件系统ceph之后，原本在系统中镜像将无法使用，所以建议当前的镜像删除，部署好ceph之后，再统一上传至ceph中存储。
2.原理解析
使用ceph的rbd接口，需要通过libvirt，所以需要在客户端机器上安装libvirt和qemu，关于ceph和openstack结合的结构如下，同时，在openstack中，需要用到存储的地方有三个:1. glance的镜像，默认的本地存储，路径在/var/lib/glance/p_w_picpaths目录下，2. nova虚拟机存储，默认本地，路径位于/var/lib/nova/instances目录下，3. cinder存储，默认采用LVM的存储方式。
3. glance与ceph联动
1.创建资源池pool
1、ceph默认创建了一个pool:rbd [root@controller_10_1_2_230 ~]# ceph osd lspools 0 rbd, [root@controller_10_1_2_230 ~]# ceph osd pool stats pool rbd id 0 nothing is going on 2、创建一个pool，指定pg_num的大小为128 [root@controller_10_1_2_230 ~]# ceph osd pool create p_w_picpaths 128 pool 'p_w_picpaths' created 3、查看pool的pg_num和pgp_num大小 [root@controller_10_1_2_230 ~]# ceph osd pool get p_w_picpaths pg_num pg_num: 128 [root@controller_10_1_2_230 ~]# ceph osd pool get p_w_picpaths pgp_num pgp_num: 128 4、查看ceph中的pools [root@controller_10_1_2_230 ~]# ceph osd lspools 0 rbd,1 p_w_picpaths, [root@controller_10_1_2_230 ~]# ceph osd pool stats pool rbd id 0 nothing is going on pool p_w_picpaths id 1 #增加了一个pool，id号码是1 nothing is going on
2.配置ceph客户端
1. glance作为ceph的客户端，即glance-api，需要有ceph的配置文件，从ceph的monitor节点复制一份配置文件过去即可，我所在环境中控制节点和ceph monitor为同一台机器，不需要操作 #如果controller节点和ceph的monitor节点是分开，则需要复制 [root@controller_10_1_2_230 ~]# scp /etc/ceph/ceph.conf root@controller_10_1_2_230:/etc/ceph/ ceph.conf 2. 安装客户端rpm包 [root@controller_10_1_2_230 ~]# yum install python-rbd -y
3.配置ceph认证
1. 添加认证的key [root@controller_10_1_2_230 ~]# ceph auth get-or-create client.glance mon 'allow r' osd 'class-read object_prefix rbd_children,allow rwx pool=p_w_picpaths' [client.glance] key = AQB526lWOOIQBxAA0ZSk30DBShtti3fKkm4aeA== 2. 查看认证列表 [root@controller_10_1_2_230 ~]# ceph auth list installed auth entries: osd.0 key: AQDsx6lWYGehDxAAGwcYP9jDvH2Zaa8JlGwj1Q== caps: [mon] allow profile osd caps: [osd] allow * osd.1 key: AQD1x6lWQCYBERAAjIKO1LVpj8FvVefDvNQZSA== caps: [mon] allow profile osd caps: [osd] allow * client.admin key: AQCexqlWQL6OGBAA2v5LsYEB5VgLyq/K2huY3A== caps: [mds] allow caps: [mon] allow * caps: [osd] allow * client.bootstrap-mds key: AQCexqlWUMNRMRAAZEp/UlhQuaixMcNy5d5pPw== caps: [mon] allow profile bootstrap-mds client.bootstrap-osd key: AQCexqlWQFfpJBAAfPCx4sTLNztBESyFKys9LQ== caps: [mon] allow profile bootstrap-osd client.glance #glance连接ceph的认证信息 key: AQB526lWOOIQBxAA0ZSk30DBShtti3fKkm4aeA== caps: [mon] allow r caps: [osd] class-read object_prefix rbd_children,allow rwx pool=p_w_picpaths 3. 将glance生成的key拷贝至 [root@controller_10_1_2_230 ~]# ceph auth get-or-create client.glance [client.glance] key = AQB526lWOOIQBxAA0ZSk30DBShtti3fKkm4aeA== #将key导出到客户端 [root@controller_10_1_2_230 ~]# ceph auth get-or-create client.glance | tee /etc/ceph/ceph.client.glance.keyring [client.glance] key = AQB526lWOOIQBxAA0ZSk30DBShtti3fKkm4aeA== [root@controller_10_1_2_230 ~]# chown glance:glance /etc/ceph/ceph.client.glance.keyring [root@controller_10_1_2_230 ~]# ll /etc/ceph/ceph.client.glance.keyring -rw-r--r-- 1 glance glance 64 Jan 28 17:17 /etc/ceph/ceph.client.glance.keyring
4. 配置glance使用ceph做为后端存储
1、备份glance-api的配置文件，以便于恢复 [root@controller_10_1_2_230 ~]# cp /etc/glance/glance-api.conf /etc/glance/glance-api.conf.orig 2、修改glance配置文件，连接至ceph [root@controller_10_1_2_230 ~]# vim /etc/glance/glance-api.conf [DEFAULT] notification_driver = messaging rabbit_hosts = 10.1.2.230:5672 rabbit_retry_interval = 1 rabbit_retry_backoff = 2 rabbit_max_retries = 0 rabbit_ha_queues = True rabbit_durable_queues = False rabbit_userid = glance rabbit_password = GLANCE_MQPASS rabbit_virtual_host = /glance default_store=rbd #glance使用的后端存储 known_stores=glance.store.rbd.Store #配置rbd的驱动 rbd_store_ceph_conf=/etc/ceph/ceph.conf #ceph的配置文件，包含有monitor的地址，通过查找monitor，可以获取认证信息 rbd_store_user=glance #认证用户，即是刚创建的用户 rbd_store_pool=p_w_picpaths #连接的存储池 rbd_store_chunk_size=8 #设置chunk size，即切割的大小 3. 重启glance服务 [root@controller_10_1_2_230 ~]# /etc/init.d/openstack-glance-api restart Stopping openstack-glance-api: [ OK ] Starting openstack-glance-api: [ OK ] [root@controller_10_1_2_230 ~]# /etc/init.d/openstack-glance-registry restart Stopping openstack-glance-registry: [ OK ] Starting openstack-glance-registry: [ OK ] [root@controller_10_1_2_230 ~]# tail -2 /etc/glance/glance-api.conf # location strategy defined by the 'location_strategy' config option. #store_type_preference = [root@controller_10_1_2_230 ~]# tail -2 /var/log/glance/registry.log 2016-01-28 18:40:25.231 21890 INFO glance.wsgi.server [-] Started child 21896 2016-01-28 18:40:25.232 21896 INFO glance.wsgi.server [-] (21896) wsgi starting up on http://0.0.0.0:9191/
5. 测试glance和ceph联动情况
[root@controller_10_1_2_230 ~]# glance --debug p_w_picpath-create --name glance_ceph_test --disk-format qcow2 --container-format bare --file cirros-0.3.3-x86_64-disk.img
curl -i -X POST -H 'x-p_w_picpath-meta-container_format: bare' -H 'Transfer-Encoding: chunked' -H 'User-Agent: python-glanceclient' -H 'x-p_w_picpath-meta-size: 13200896' -H 'x-p_w_picpath-meta-is_public: False' -H 'X-Auth-Token: 062af9027a85487997d176c9f1e963f2' -H 'Content-Type: application/octet-stream' -H 'x-p_w_picpath-meta-disk_format: qcow2' -H 'x-p_w_picpath-meta-name: glance_ceph_test' -d '<open file u'cirros-0.3.3-x86_64-disk.img', mode 'rb' at 0x1ba24b0>' http://controller:9292/v1/p_w_picpaths
HTTP/1.1 201 Created
content-length: 489
etag: 133eae9fb1c98f45894a4e60d8736619
location: http://controller:9292/v1/p_w_picpaths/348a90e8-3631-4a66-a45d-590ec6413e7d
date: Thu, 28 Jan 2016 10:42:06 GMT
content-type: application/json
x-openstack-request-id: req-b993bc0b-447e-49b4-a8ce-bd7765199d5a
{"p_w_picpath": {"status": "active", "deleted": false, "container_format": "bare", "min_ram": 0, "updated_at": "2016-01-28T10:42:06", "owner": "ef4b83a909dc4689b663ff2c70022478", "min_disk": 0, "is_public": false, "deleted_at": null, "id": "348a90e8-3631-4a66-a45d-590ec6413e7d", "size": 13200896, "virtual_size": null, "name": "glance_ceph_test", "checksum": "133eae9fb1c98f45894a4e60d8736619", "created_at": "2016-01-28T10:42:04", "disk_format": "qcow2", "properties": {}, "protected": false}}
+------------------+--------------------------------------+
| Property | Value |
+------------------+--------------------------------------+
| checksum | 133eae9fb1c98f45894a4e60d8736619 |
| container_format | bare |
| created_at | 2016-01-28T10:42:04 |
| deleted | False |
| deleted_at | None |
| disk_format | qcow2 |
| id | 348a90e8-3631-4a66-a45d-590ec6413e7d |
| is_public | False |
| min_disk | 0 |
| min_ram | 0 |
| name | glance_ceph_test |
| owner | ef4b83a909dc4689b663ff2c70022478 |
| protected | False |
| size | 13200896 |
| status | active |
| updated_at | 2016-01-28T10:42:06 |
| virtual_size | None |
+------------------+--------------------------------------+
[root@controller_10_1_2_230 ~]# glance p_w_picpath-list
+--------------------------------------+---------------------+-------------+------------------+----------+--------+
| ID | Name | Disk Format | Container Format | Size | Status |
+--------------------------------------+---------------------+-------------+------------------+----------+--------+
| 56e96957-1308-45c7-9c66-1afff680b217 | cirros-0.3.3-x86_64 | qcow2 | bare | 13200896 | active |
| 348a90e8-3631-4a66-a45d-590ec6413e7d | glance_ceph_test | qcow2 | bare | 13200896 | active | #上传成功
+--------------------------------------+---------------------+-------------+------------------+----------+--------+6.查看ceph池的数据
[root@controller_10_1_2_230 ~]# rados -p p_w_picpaths ls rbd_directory rbd_header.10d7caaf292 rbd_data.10dd1fd73446.0000000000000001 rbd_id.348a90e8-3631-4a66-a45d-590ec6413e7d rbd_header.10dd1fd73446 rbd_data.10d7caaf292.0000000000000000 rbd_data.10dd1fd73446.0000000000000000 rbd_id.8a09b280-5916-44c6-9ce8-33bb57a09dad @@@glance中的数据存储到了ceph文件系统中@@@
4. 总结
将openstack的glance的数据存储到ceph中是一种非常好的解决方案，既能够保障p_w_picpath数据的安全性，同时glance和nova在同个存储池中，能够基于copy-on-write的方式快速创建虚拟机，能够在秒级为单位实现vm的创建。
