rexray在CentOS上不能创建ceph rbd的docker volume问题定位

背景

  我们通过docker的rexray插件来创建ceph rbd设备的docker volume,但总提示创建失败。

# docker volume create --driver=rexray --opt=size=5 --name=cephrbd-book
Error response from daemon: create test_cephrbd_volume: VolumeDriver.Create: {"Error":"Failed to create new volume"}

  OS版本:

# lsb_release -a
LSB Version:	:core-4.1-amd64:core-4.1-noarch
Distributor ID:	CentOS
Description:	CentOS Linux release 7.4.1708 (Core) 
Release:	    7.4.1708
Codename:   	Core

  Ceph版本:

# ceph version
ceph version 0.94.5

  rexray版本:

# rexray version
REX-Ray
-------
Binary: /usr/bin/rexray
Flavor: client+agent+controller
SemVer: 0.9.0
OsArch: Linux-x86_64
Branch: (detached from 2a7458d
Commit: 2a7458dd90a79c673463e14094377baf9fc8695e
Formed: Wed, 26 Jul 2017 14:35:37 CST

libStorage
----------
SemVer: 0.6.0
OsArch: Linux-x86_64
Branch: (detached from fa055d6
Commit: fa055d6da595602715bdfd5541b4aa6d4dcbcbd9
Formed: Wed, 26 Jul 2017 14:35:11 CST

分析

  之前我们在ubuntu 16.04的环境中,也搭建过docker+ceph+rexray的环境,使用docker命令创建volume并没有报错。正常的流程是:

1. 创建docker volume:rexray会使用rbd create命令创建一个rbd设备;
2. 使用该volume创建docker 容器:rexray会将rbd 设备map到docker host上成为一个/dev/rbd设备,再使用mount命令将/dev/rbd设备挂载到/var/lib/libstorage/volumes/目录,供容器使用;
3. 删除该容器时:先umount掉/dev/rbd设备在/var/lib/libstorage/volumes/目录的挂载,再rbd unmap掉rbd设备在docker host上的映射;
4. 删除docker volume:rexray会使用rbd rm命令删除该rbd设备;

  在我们的实际使用情况中,发现只有第一步会出错,后面的三步都并未出现任何异常。而且,如果我们不使用docker volume create命令创建rbd设备,而是直接使用“rbd create”命令来创建rbd设备,也并未出现任何错误,且该rbd设备也可以正常使用。这是很奇怪的一个地方。
  查看docker日志,也只是提示“Failed to create new volume”,并没有其他有用的信息。

# journalctl -xu docker
...
Jan 22 10:45:13 dcos-agent2 dockerd[66886]: time="2018-01-22T10:45:13.937645601+08:00" level=error msg="Handler for POST /v1.29/volumes/create returned error: create cephrbd-book: VolumeDri
ver.Create: {"Error":"Failed to create new volume"}
"

  既然docker中没有更有价值的日志,那么就尝试看下rexray的日志,这里需要打开rexray的debug选项(打开方法参考),从而可以看到更多的日志。

# vim /var/log/rexray/rexray.log
...
time="2018-01-19T19:12:32+08:00" level=info msg="    -------------------------- HTTP REQUEST (CLIENT) -------------------------"
time="2018-01-19T19:12:32+08:00" level=info msg="    GET /volumes/rbd?attachments=0 HTTP/1.1"
time="2018-01-19T19:12:32+08:00" level=info msg="    Host: libstorage-server"
time="2018-01-19T19:12:32+08:00" level=info msg="    Libstorage-Instanceid: rbd=109.105.115.73"
time="2018-01-19T19:12:32+08:00" level=info msg="    Libstorage-Localdevices: rbd="
time="2018-01-19T19:12:32+08:00" level=info msg="    Libstorage-Tx: txID=b9fb044b-d47a-4ac5-7533-5824b8dc737a, txCR=1516360352"
time="2018-01-19T19:12:32+08:00" level=info msg="    "
time="2018-01-19T19:12:33+08:00" level=info
time="2018-01-19T19:12:33+08:00" level=info msg="    -------------------------- HTTP RESPONSE (CLIENT) -------------------------"
time="2018-01-19T19:12:33+08:00" level=info msg="    HTTP/1.1 200 OK"
time="2018-01-19T19:12:33+08:00" level=info msg="    Content-Length: 228"
time="2018-01-19T19:12:33+08:00" level=info msg="    Content-Type: application/json"
time="2018-01-19T19:12:33+08:00" level=info msg="    Date: Fri, 19 Jan 2018 11:12:33 GMT"
time="2018-01-19T19:12:33+08:00" level=info msg="    Libstorage-Servername: jade-chopper-ky"
time="2018-01-19T19:12:33+08:00" level=info msg="    "
time="2018-01-19T19:12:33+08:00" level=info msg="    {"
time="2018-01-19T19:12:33+08:00" level=info msg="      "rbd.test_fs_device2": {"
time="2018-01-19T19:12:33+08:00" level=info msg="        "name": "test_fs_device2","
time="2018-01-19T19:12:33+08:00" level=info msg="        "size": 16,"
time="2018-01-19T19:12:33+08:00" level=info msg="        "id": "rbd.test_fs_device2","
time="2018-01-19T19:12:33+08:00" level=info msg="        "type": "rbd""
time="2018-01-19T19:12:33+08:00" level=info msg="      },"
time="2018-01-19T19:12:33+08:00" level=info msg="      "rbd.test_majk": {"
time="2018-01-19T19:12:33+08:00" level=info msg="        "name": "test_majk","
time="2018-01-19T19:12:33+08:00" level=info msg="        "id": "rbd.test_majk","
time="2018-01-19T19:12:33+08:00" level=info msg="        "type": "rbd""
time="2018-01-19T19:12:33+08:00" level=info msg="      }"
time="2018-01-19T19:12:33+08:00" level=info msg="    }"
time="2018-01-19T19:12:33+08:00" level=info
time="2018-01-19T19:12:33+08:00" level=info msg="    -------------------------- HTTP REQUEST (CLIENT) -------------------------"
time="2018-01-19T19:12:33+08:00" level=info msg="    POST /volumes/rbd HTTP/1.1"
time="2018-01-19T19:12:33+08:00" level=info msg="    Host: libstorage-server"
time="2018-01-19T19:12:33+08:00" level=info msg="    Libstorage-Instanceid: rbd=109.105.115.73"
time="2018-01-19T19:12:33+08:00" level=info msg="    Libstorage-Localdevices: rbd="
time="2018-01-19T19:12:33+08:00" level=info msg="    Libstorage-Tx: txID=b1565614-1b5f-4de2-5756-b74fb99887aa, txCR=1516360353"
time="2018-01-19T19:12:33+08:00" level=info msg="    "
time="2018-01-19T19:12:33+08:00" level=info msg="    {"name":"cephrbd-book","availabilityZone":"","iops":0,"size":5,"type":"","opts":{"size":"5"}}"
time="2018-01-19T19:12:33+08:00" level=info
time="2018-01-19T19:12:33+08:00" level=info msg="    -------------------------- HTTP RESPONSE (CLIENT) -------------------------"
time="2018-01-19T19:12:33+08:00" level=info msg="    HTTP/1.1 500 Internal Server Error"
time="2018-01-19T19:12:33+08:00" level=info msg="    Content-Length: 319"
time="2018-01-19T19:12:33+08:00" level=info msg="    Content-Type: application/json"
time="2018-01-19T19:12:33+08:00" level=info msg="    Date: Fri, 19 Jan 2018 11:12:33 GMT"
time="2018-01-19T19:12:33+08:00" level=info msg="    Libstorage-Servername: jade-chopper-ky"
time="2018-01-19T19:12:33+08:00" level=info msg="    "
time="2018-01-19T19:12:33+08:00" level=info msg="    {"
time="2018-01-19T19:12:33+08:00" level=info msg="      "message": "Failed to create new volume","
time="2018-01-19T19:12:33+08:00" level=info msg="      "status": 500,"
time="2018-01-19T19:12:33+08:00" level=info msg="      "error": {"
time="2018-01-19T19:12:33+08:00" level=info msg="        "driverName": "rbd","
time="2018-01-19T19:12:33+08:00" level=info msg="        "inner": {"
time="2018-01-19T19:12:33+08:00" level=info msg="          "inner": "Error running command: [rbd: strict_strtoll: garbage at end of string. got: '5G'\n]","
time="2018-01-19T19:12:33+08:00" level=info msg="          "msg": "unable to create rbd""
time="2018-01-19T19:12:33+08:00" level=info msg="        },"
time="2018-01-19T19:12:33+08:00" level=info msg="        "opts.Size": 5,"
time="2018-01-19T19:12:33+08:00" level=info msg="        "volumeName": "cephrbd-book""
time="2018-01-19T19:12:33+08:00" level=info msg="      }"
time="2018-01-19T19:12:33+08:00" level=info msg="    }"

  通过日志可以看到,每次发送docker volume create命令时,实际发送了两次http请求,第一次是列出当前的rbd设备,第二次才是发出post请求来创建一个rbd设备。而这里的日志,则给出了更多的信息“rbd: strict_strtoll: garbage at end of string. got: '5G'”。也就是说,很可能就是在rexray调用rbd create命令时发生了错误。这里,我们打开rexray源码继续跟踪,具体的源码文件为rexray/blob/master/libstorage/drivers/storage/rbd/utils/utils.go。

//RBDCreate creates a new RBD volume on the cluster
func RBDCreate(
	ctx types.Context,
	pool *string,
	image *string,
	sizeGB *int64,
	objectSize *string,
	features []*string) error {

	cmd := exec.Command(
		rbdCmd, "create", poolOpt, *pool,
		"--object-size", *objectSize,
		"--size", strconv.FormatInt(*sizeGB, 10)+"G",
	)

	for _, feature := range features {
		cmd.Args = append(cmd.Args, "--image-feature")
		cmd.Args = append(cmd.Args, *feature)
	}

	cmd.Args = append(cmd.Args, *image)
	_, _, err := RunCommand(ctx, cmd)
	if err != nil {
		return goof.WithError("unable to create rbd", err)
	}

	return nil
}

  根据上面的代码可以大致看出,rexray就是直接调用rbd create命令来创建rbd设备,和我们直接调用命令不同的是它这里是把rbd的大小转化成GB,然后再字符串拼接了一个“G”。这里我们就按照它这里的方法,组装了一个新的rbd命令,然后在CentOS上测试:

# rbd create cephrbd-test --size 5G
rbd: strict_strtoll: garbage at end of string. got: '5G'

  果然,还是报错了,我们再使用该命令在ubuntu上测试:

# rbd create cephrbd-test --size 5G

  在ubuntu上运行,没有报错。同一条命令,在不同的环境下结果不同。那么最大的可能就是rbd(ceph)的版本不一致了。在ubuntu下获取ceph版本。

# ceph version
ceph version 10.2.7

  而CentOS上默认安装的ceph版本只是0.94.5(Hammer版),而Ubuntu上默认安装的却是更高的10.2.7(jewel版本)。而0.94.5 ceph版本的rbd命令无法解析“G”,“M”等描述rbd大小的参数。这就是问题出现的原因。

解决方法

  在CentOS上安装更新的Ceph版本(jewel或更新的版本)即可。

ps:后面的博客会介绍如何在CentOS上安装更新的Ceph版本。

原文地址:https://www.cnblogs.com/styshoo/p/8339811.html