OpenStack 存储：Ceph 与 GlusterFS

Question

目前正在将 OpenStack 部署到本地计算集群，但使用OpenStack Ansible Playbook我似乎在 setup-infrastructure.yml 部署的基础架构上遇到了问题。

之前提过 Ceph Storage 集群，但基础设施脚本在尝试部署 glusterfs 时崩溃。

<hideko_repo_container-b0e9b1e5> Task "Create gluster peers" has been omitted from the job because the conditional "['_glusterfs_is_last_play_host']" was evaluated as "False"
<nusselt_repo_container-b7d3e224> Task "Create gluster peers" has been omitted from the job because the conditional "['_glusterfs_is_last_play_host']" was evaluated as "False"
<kushana_repo_container-5ce108f3> Task "Create gluster peers" has been omitted from the job because the conditional "['_glusterfs_is_last_play_host']" was evaluated as "False"
<ursell_repo_container-823f4013> Task "Create gluster peers" has been omitted from the job because the conditional "['_glusterfs_is_last_play_host']" was evaluated as "False"
<lundquist_repo_container-f1b53511> The "physical_host" variable of "lundquist" has been found to have a corresponding host entry in inventory.
<lundquist_repo_container-f1b53511> The "physical_host" variable of "lundquist" terminates at "XXX.XX.XXX.XX" using the host variable "ansible_host".

然后导致以下错误：

fatal: [lundquist_repo_container-f1b53511]: FAILED! => {
    "changed": false,
    "invocation": {
        "module_args": {
            "force": null,
            "nodes": [
                "hideko-repo-container-b0e9b1e5",
                "nusselt-repo-container-b7d3e224",
                "kushana-repo-container-5ce108f3",
                "ursell-repo-container-823f4013",
                "lundquist-repo-container-f1b53511"
            ],
            "state": "present"
        }
    },
    "msg": "",
    "rc": 1
}

如果它有帮助，稍后在脚本中它再次失败并出现以下错误：

<XXX.XX.XXX.XX> Failed to connect to the host via ssh: 
The full traceback is:
NoneType: None
fatal: [ursell_repo_container-823f4013]: FAILED! => {
    "changed": false,
    "invocation": {
        "module_args": {
            "arbiters": null,
            "bricks": "/gluster/bricks/1",
            "cluster": [
                "hideko-repo-container-b0e9b1e5",
                "nusselt-repo-container-b7d3e224",
                "kushana-repo-container-5ce108f3",
                "ursell-repo-container-823f4013",
                "lundquist-repo-container-f1b53511"
            ],
            "directory": null,
            "disperses": null,
            "force": true,
            "host": null,
            "name": "gfs-repo",
            "options": {},
            "quota": null,
            "rebalance": false,
            "redundancies": null,
            "replicas": 5,
            "start_on_create": true,
            "state": "present",
            "stripes": null,
            "transport": "tcp"
        }
    },
    "msg": "error running gluster (/usr/sbin/gluster --mode=script peer probe nusselt-repo-container-b7d3e224) command (rc=1): Error : Request timed out\n"
}

是否可能取决于 ssh 连接的工作方式？ 现在只有头节点（Hideko）可以无密码连接到其他节点，但其他节点不能以相同的方式连接回来。

当我有 Ceph 集群时，我真的需要 glusterfs 吗？ 从我的（世俗的）理解来看，这两者是相互冲突的？

非常感谢任何指针:))

• 路德格