Openstack上使用BOSH部署CloudFoundry（二）

第二篇 ：部署BOSH

micro bosh部署完毕后，要利用micro bosh部署bosh，先从bosh-cli target到micro bosh上

ubuntu@bosh-cli:~/deployments$ bosh target 10.23.51.26:25555
Target already set to `microbosh-openstack'
ubuntu@bosh-cli:~/deployments$ bosh login
Your username: admin
Enter password: *****
Logged in as `admin'

默认用户密码为admin admin

ubuntu@bosh-cli:~/deployments$ bosh deployments

发现还没有部署任何项目。那么，下面就着手一步步部署bosh，本文只专注部署过程以及碰到的问题，bosh模块简介等不涉及。

BOSH stemcell

部署micro bosh 用的是mirco bosh stemcell，而部署bosh则用 bosh stemcell，两个stemcell的区别在上一篇中已经提到。

bosh public stemcells --all

选择openstack 最新版本

bosh download public stemcell bosh-stemcell-openstack-kvm-0.7.0.tgz

上传bosh stemcell

bosh upload stemcell bosh-stemcell-openstack-kvm-0.7.0.tgz

bosh-release编译包

先从github或自己的代码库中取最新版本，这里是用github上的代码。

git clone https://github.com/cloudfoundry/bosh.git[/code] 
git submodule update --init –recursive


bosh create release –with-tarball --force


不出情况的话，会生成release包。这一步得看网速和网络环境了，rubygem你懂的。有快有慢，如果有提示哪个模块没有，其实可以进入文件夹，手工去git clone。

如中间碰到某些gem搞不下来， 那是rubygem问题了，多多重试。最不济就用goagent，虽然在micro bosh上再搞个goagent有点别扭。

...
...
Release version:11.1-dev
Release manifest:/home/ubuntu/bosh-release/dev_releases/bosh-dev1-11.1-dev.yml
Release tarball (88.8M):/home/ubuntu/bosh-release/dev_releases/bosh-dev1-11.1-dev.tgz


上传bosh-release

bosh upload release ~/bosh-release/dev_releases/bosh-dev-11.1-dev.tgz


编写bosh部署文件bosh.yml
bosh.yml 定义了安装的模块，以及IP、内存等信息

---
name: bosh-openstack #bosh名称，部署完bosh deployments列出来的就是
director_uuid: f6d5ef18-441d-40a4-b895-0a49f72097d4 #bosh status ,把uudi copy过来

release:
name: bosh-dev   # bosh releases,查看上传的release,name和version要对应上
version: 11.1-dev

compilation:
workers: 1
network: default
reuse_compilation_vms: true
cloud_properties:
instance_type: jae.bosh # 此处的flavor要自己提前定义

update:
canaries: 1
canary_watch_time: 90000
update_watch_time: 90000
max_in_flight: 1
max_errors: 1

networks:
- name: floating
type: vip
cloud_properties: {}
- name: default
type: dynamic
dns:
- 10.23.54.108 # CHANGE: Micro Bosh IP
cloud_properties:
security_groups:
- default

resource_pools:
- name: small
network: default
size: 8
stemcell:
name: bosh-stemcell
version: 0.7.0
cloud_properties:
instance_type: m3.small-no-delete #自定义flavor
env:
bosh:
password: $6$D0NatspPe$CscRzCu.SHVOBW0V114PyfXbESPLbOmx9TpHGEWKAYtGD3IXDCqUzdfinyjWTMxccder7AFuy0Z5mej2yzR5f.

jobs:
- name: powerdns
template: powerdns
instances: 1
resource_pool: small
networks:
- name: default
default: [dns, gateway]
- name: floating
static_ips:
- 10.23.54.121 # Change: Floating IP 1

- name: nats
template: nats
instances: 1
resource_pool: small
networks:
- name: default
default: [dns, gateway]

- name: postgres
template: postgres
instances: 1
resource_pool: small
persistent_disk: 2048
networks:
- name: default
default: [dns, gateway]

- name: redis
template: redis
instances: 1
resource_pool: small
networks:
- name: default
default: [dns, gateway]

- name: director
template: director
instances: 1
resource_pool: small
persistent_disk: 4096
networks:
- name: default
default: [dns, gateway]
- name: floating
static_ips:
- 10.23.54.122 # Change: Floating IP 2

- name: blobstore
template: blobstore
instances: 1
resource_pool: small
persistent_disk: 8192
networks:
- name: default
default: [dns, gateway]

- name: openstack_registry
template: openstack_registry
instances: 1
resource_pool: small
networks:
- name: default
default: [dns, gateway]
- name: floating
static_ips:
- 10.23.54.126 # Change: Floating IP 3

- name: health_monitor
template: health_monitor
instances: 1
resource_pool: small
networks:
- name: default
default: [dns, gateway]

properties:
env:

dns:
user: powerdns
password: powerdns
address: 10.23.54.121 # Change: Floating IP 1
replication:
basic_auth: replication:QRoqsenPsNGX6
user: replication
password: powerdns
database:
name: powerdns
port: 5432
#recursors:
#  - 10.23.51.10
webserver:
password: powerdns
port: 8081

nats:
user: nats
password: nats
address:  0.nats.default.bosh-openstack.bosh
listen_address: 0.0.0.0
port: 4222

postgres:
user: postgres
password: postgres
address: 0.postgres.default.bosh-openstack.bosh
listen_address: 0.0.0.0
port: 5432
database: bosh

redis:
address: 0.redis.default.bosh-openstack.bosh
listen_address: 0.0.0.0
port: 25255
password: redis

director:
name: bosh-openstack
address: 10.23.54.122 # Change: Floating IP 2
port: 25555
encryption: false

blobstore:
address: 0.blobstore.default.bosh-openstack.bosh
port: 25250
backend_port: 25251
agent:
user: agent
password: agent
director:
user: director
password: director

openstack_registry:
address: 10.40.62.126  # Change: Floating IP 3
http:
port: 25777
user: registry
password: registry
db:
max_connections: 32
pool_timeout: 10

hm:
http:
port: 25923
user: hm
password: hm
director_account:
user: admin
password: admin
intervals:
prune_events: 30
poll_director: 60
poll_grace_period: 30
log_stats: 60
analyze_agents: 60
agent_timeout: 180
rogue_agent_alert: 180
loglevel: info
email_notifications: false
tsdb_enabled: false
pagerduty_enable: false
varz_enabled: false

openstack:
auth_url: http://ip:5000/v2.0/tokens # CHANGE: OpenStack Auth URL
username: xx # CHANGE: OpenStack Username
api_key: xx # CHANGE: OpenStack Password
tenant: jae # CHANGE: OpenStack Tenant
default_key_name: xx2 # CHANGE: OpenStack default Keyname
default_security_groups: ["default"] # CHANGE: OpenStack default Security Group
private_key: ~/.ssh/xx2.pem # CHANGE: OpenStack Private Key


部署bosh
ubuntu@bosh-cli:~/deployments$ bosh deployment bosh.yml
Deployment set to `/home/ubuntu/deployments/bosh-openstack/bosh.yml'
ubuntu@bosh-cli:~/deployments$ bosh deploy


部署成功后，bosh deployments 和 bosh vms查看部署情况。

ubuntu@bosh-cli:~/deployments$ bosh deployments

+----------------+-------------------+---------------------+
| Name           | Release(s)        | Stemcell(s)         |
+----------------+-------------------+---------------------+
| bosh-openstack | bosh-dev/11.1-dev | bosh-stemcell/0.7.0 |
+----------------+-------------------+---------------------+

Deployments total: 1
ubuntu@bosh-cli:~/deployments$ bosh vms
Deployment `bosh-openstack'

Director task 33

Task 33 done

+----------------------+---------+---------------+---------------------------+
| Job/index            | State   | Resource Pool | IPs                       |
+----------------------+---------+---------------+---------------------------+
| blobstore/0          | running | small         | 10.23.51.71               |
| director/0           | running | small         | 10.23.51.69, 10.23.54.122 |
| health_monitor/0     | running | small         | 10.23.51.80               |
| nats/0               | running | small         | 10.23.51.79               |
| openstack_registry/0 | running | small         | 10.23.51.77, 10.23.54.126 |
| postgres/0           | running | small         | 10.23.51.73               |
| powerdns/0           | running | small         | 10.23.51.41, 10.23.54.121 |
| redis/0              | running | small         | 10.23.51.66               |
+----------------------+---------+---------------+---------------------------+

VMs total: 8


遇到的问题
其实部署bosh，官网等都有详细的文档。但是过程中的问题总是神出鬼没，所以这一个环节显得很有必要。

问题一：openstack用户的quota配额
当执行bosh deploy的时候，问题就集中出现了，第一个碰到的问题是openstack用户的quota配额问题。

Creating bound missing VMs
small/2: Expected([200, 202]) <=> Actual(413 Request Entity Too Large)
request => {:connect_timeout=>60, :headers=>{"Content-Type"=>"application/json", "X-Auth-Token"=>"38fe51b931184a30a287e71bc37cc05d", "Host"=>"10.23.54.150:8774", "Content-Length"=>422}, :instrumentor_name=>"excon", :mock=>false, :nonblock=>true, :read_timeout=>60, :retry_limit=>4, :ssl_ca_file=>"/var/vcap/packages/director/bosh/director/vendor/bundle/ruby/1.9.1/gems/excon-0.16.2/data/cacert.pem", :ssl_verify_peer=>true, :write_timeout=>60, :host=>"10.23.54.150", :path=>"/v2/69816bacecd749f9ba1d68b3c8bae1f1/servers.json", :port=>"8774", :query=>"ignore_awful_caching1362453746", :scheme=>"http", :body=>"{\"server\":{\"flavorRef\":\"25\",\"imageRef\":\"e205b9ec-0e19-4500-87fe-ede3af13b227\",\"name\":\"vm-b875d6d8-81ce-483b-bfa8-d6d525aaf280\",\"metadata\":{},\"user_data\":\"eyJyZWdpc3RyeSI6eyJlbmRwb2ludCI6Imh0dHA6Ly8xMC4yMy41MS4zNToy\\nNTc3NyJ9LCJzZXJ2ZXIiOnsibmFtZSI6InZtLWI4NzVkNmQ4LTgxY2UtNDgz\\nYi1iZmE4LWQ2ZDUyNWFhZjI4MCJ9LCJkbnMiOnsibmFtZXNlcnZlciI6WyIx\\nMC4yMy41NC4xMDgiXX19\\n\",\"key_name\":\"jae2\",\"security_groups\":[{\"name\":\"default\"}]}}", :expects=>[200, 202], :method=>"POST"}
response => #<Excon::Response:0x00000004edec30 @body="{\"overLimit\": {\"message\": \"Quota exceeded for instances: Requested 1, but already used 10 of 10 instances\", \"code\": 413}}", @headers={"Retry-After"=>"0", "Content-Length"=>"121", "Content-Type"=>"application/json; charset=UTF-8", "X-Compute-Request-Id"=>"req-c5427ed2-62af-47b9-98a6-6f114893d8fc", "Date"=>"Tue, 05 Mar 2013 03:22:27 GMT"}, @status=413> (00:00:02)


看response部分：

response => #<Excon::Response:0x00000004edec30 @body="{\"overLimit\": {\"message\": \"Quota exceeded for instances: Requested 1, but already used 10 of 10 instances\", \"code\": 413}}"


由于quota，不能再创建instance了。但是bosh部署的时候要创建8个组件vm + n 个workers。后面还有cf-release的几十个vm呢，必须要调整。通过openstack dashboard 修改此用户的各种配额，包括instance数，vcpu数，mem，volume等，尽量够大，省得再改麻烦。

改后再bosh deply，如果想从头再来，可以删除部署过的半拉子项目：

bosh deployments

bosh delete deployment {depoly-name}


这个命令会把创建过的vm，volume全删掉。另外要记得删掉bosh当前目录有个缓存的yml。不然会出现vm找不到的情况，因为缓存yml已经记录了创建过的vm，你删掉了可不就找不着了。

问题二：创建vm的线程问题。
Director task 22

Preparing deployment
binding deployment (00:00:00)
binding releases (00:00:00)
binding existing deployment: VM `413db010-cb45-45fb-97f5-819cd4f4b6fe' is out of sync: it reports itself as `blobstore/0' but there is no instance reference in DB (00:00:00)
Error                   3/9 00:00:00

Error 400004: VM `413db010-cb45-45fb-97f5-819cd4f4b6fe' is out of sync: it reports itself as `blobstore/0' but there is no instance reference in DB


这个问题google良久，没有找到可靠的办法。是跟创建vm的最大线程有关。由于本文的bosh-release用的是年前编译的，没有更新。查看代码，创建vm的最大线程数为32，

于是手动改了，新代码里写个变量了，可以create bosh release前改下配置。部署bosh是由micro bosh的direct统一调度的，所以要先 ssh到micro bosh去修改源码。

ssh vcap@10.23.51.26


密码为admin，在micro_bosh.yml设置的。上一篇中有提到：Openstack上使用BOSH部署CloudFoundry（一）

代码位于/var/vcap/packages/director/bosh/director/lib/director/jobs/update_deployment.rb的85行左右。新版本上已经改为变量了：

https://github.com/cloudfoundry/bosh/blob/master/director/lib/director/jobs/update_deployment.rb#L85


我是把这一行改成:

ThreadPool.new(:max_threads => 64).wrap do |thread_pool|


然后重启micro bosh 的director

root@bm-c3ffd8f1-46be-4936-becc-03c426317ade:sudo su -
root@bm-c3ffd8f1-46be-4936-becc-03c426317ade:/var/vcap/bosh/bin# ./monit restart director


另外，担心worker过多也会造成问题，直接把worker减到1。

问题三：Error 400007: `director/0' is not running after update
ubuntu@bosh-cli:~/deployments$ bosh vms
Deployment `bosh-openstack'

Director task 29

Task 29 done

+----------------------+---------+---------------+---------------------------+
| Job/index            | State   | Resource Pool | IPs                       |
+----------------------+---------+---------------+---------------------------+
| blobstore/0          | running | small         | 10.23.51.71               |
| director/0           | failing | small         | 10.23.51.69, 10.23.54.122 |
| health_monitor/0     | running | small         | 10.23.51.80               |
| nats/0               | running | small         | 10.23.51.79               |
| openstack_registry/0 | running | small         | 10.23.51.77, 10.23.54.126 |
| postgres/0           | running | small         | 10.23.51.73               |
| powerdns/0           | running | small         | 10.23.51.41, 10.23.54.121 |
| redis/0              | running | small         | 10.23.51.66               |
+----------------------+---------+---------------+---------------------------+


果然director 的状态是failing，进去openstack dashboard，手动重启下对应的vm，然后再次bosh deploy，又碰到一两次其他模块的这个问题，一样的方法解决了。这个方法
比较旁门，还是得把问题根源解决。

本文到此结束，有不足不处望指正，多多交流，微博：http://weibo.com/kingjames3

下篇讲解如何用bosh来部署cloudfoundry。