Article catalogue
1. NFS introduction
NFS is the abbreviation of network file system. As its name suggests, it is the network file storage system. It was first developed by Sun company. It is also one of the file systems supported by FreeBSD. It allows computers in the network to share resources through TCP/IP network. Through NFS, our local NFS client applications can transparently read and write files on the server NFS server, just as convenient as accessing local files. In short, NFS is a service that allows different hosts and different operating systems to share storage through the network.
In the process of file transfer or information transfer, NFS relies on RPC (Remote Procedure Call) protocol, that is, Remote Procedure Call. All functions of NFS must be registered with RPC. In this way, RPC can understand the functions of NFS, such as Port, PID, IP monitored by NFS on the server, and the client can find the correct corresponding Port through RPC query. Therefore, NFS can successfully provide services only when RPC exists. Simply understand the relationship between NFS and RPC: NFS is a file storage system, and RPC is responsible for information transmission.
2. Environment and software preparation
In this demonstration environment, I install Linux system on the virtual machine to perform the operation. The following is the installed software and version:
- Oracle VirtualBox: 5.1.20 r114628 (Qt5.6.2)
- System: CentOS Linux release 7.3.1611 (Core)
- rpcbind: 0.2.0-38.el7.x86_64
- nfs-utils: 1.3.0-0.54.el7.x86_64
3. Services for NFS installation
Through the above brief introduction, we know that service for NFS needs to rely on RPC service, so here, the NFS server needs to install rpcbind and NFS utils, and the client only needs to install NFS utils. Since I choose CentOS system, I can use yum to install quickly.
First, confirm whether the lower server system has NFS installed.
$ rpm -qa nfs-utils rpcbind nfs-utils-1.3.0-0.54.el7.x86_64 rpcbind-0.2.0-38.el7.x86_64
Note: I have completed the installation here. If it is empty, it indicates that it is not installed.
Then, install services for NFS
# Server $ yum install -y nfs-utils rpcbind # client $ yum install -y nfs-utils
Another installation command for Ubuntu 16.04 is:
# Server apt install nfs-kernel-server # client apt install nfs-common
4. NFS configuration and use
We create a shared directory / data/share on the server side as the remote entry for the client to mount, and then set the permissions.
$ mkdir -p /data/share $ chmod 666 /data/share
Then, modify the NFS configuration file / etc/exports
$ vim /etc/exports /data/share 10.222.77.0/24(rw,sync,insecure,no_subtree_check,no_root_squash)
To explain, there are many parameters behind the configuration here, and each parameter has different meanings. For details, please refer to the following. Here, I configured the / data/share file directory to allow clients with IP in the 10.222.77.0/24 range to mount. Of course, if the client IP is not in the range and wants to mount, you can set the IP range to be larger or set it to * to allow all clients to mount. For example, / home * (RO, sync, secure, no_root_square) set the / home directory to allow all clients to mount read-only.
| parameter | explain |
|---|---|
| ro | read-only access |
| rw | Read write access |
| sync | All data is written to the share on request |
| async | nfs can respond to requests before writing data |
| secure | nfs is sent through secure TCP/IP ports below 1024 |
| insecure | nfs is sent through ports above 1024 |
| wdelay | If multiple users want to write to the nfs directory, write as a group (default) |
| no_wdelay | If multiple users want to write to the nfs directory, write immediately. This setting is not required when async is used |
| hide | Do not share its subdirectories in the nfs shared directory |
| no_hide | Subdirectories of shared nfs directory |
| subtree_check | When sharing subdirectories such as / usr/bin, force nfs to check the permissions of the parent directory (default) |
| no_subtree_check | Do not check parent directory permissions |
| all_squash | UID and GID of shared file map anonymous user anonymous, which is suitable for public directory |
| no_all_squash | Keep UID and GID of shared files (default) |
| root_squash | All requests of the root user are mapped to the same permissions as the anonymous user (default) |
| no_root_squash | The root user has full administrative access to the root directory |
| anonuid=xxx | Specifies the UID of the anonymous user in the / etc/passwd file of the nfs server |
| anongid=xxx | Specifies the GID of the anonymous user in the / etc/passwd file of the nfs server |
Next, let's start the RPC service.
$ service rpcbind start
# Or you can use the following command
$ /bin/systemctl start rpcbind.service
# View the list of ports registered by rpc server of NFS service item
$ rpcinfo -p localhost
program vers proto port service
100000 4 tcp 111 portmapper
100000 3 tcp 111 portmapper
100000 2 tcp 111 portmapper
100000 4 udp 111 portmapper
100000 3 udp 111 portmapper
Now, let's take a look at the list of NFS service 111 ports that haven't been started. Let's just look at the list of NFS service 111 ports that haven't been started.
# Start NFS service
$ service nfs start
# Or you can use the following command
/bin/systemctl start nfs.service
# After starting service for NFS, rpc service has enabled port mapping list for NFS
# rpcinfo -p localhost
program vers proto port service
100000 4 tcp 111 portmapper
100000 3 tcp 111 portmapper
100000 2 tcp 111 portmapper
100000 4 udp 111 portmapper
100000 3 udp 111 portmapper
100000 2 udp 111 portmapper
100024 1 udp 33745 status
100024 1 tcp 36980 status
100005 1 udp 20048 mountd
100005 1 tcp 20048 mountd
100005 2 udp 20048 mountd
100005 2 tcp 20048 mountd
100005 3 udp 20048 mountd
100005 3 tcp 20048 mountd
100003 3 tcp 2049 nfs
100003 4 tcp 2049 nfs
100227 3 tcp 2049 nfs_acl
100003 3 udp 2049 nfs
100003 4 udp 2049 nfs
100227 3 udp 2049 nfs_acl
100021 1 udp 38960 nlockmgr
100021 3 udp 38960 nlockmgr
100021 4 udp 38960 nlockmgr
100021 1 tcp 38362 nlockmgr
100021 3 tcp 38362 nlockmgr
100021 4 tcp 38362 nlockmgr
We found that after the NFS service was started, the port list registered by rpc increased significantly. OK now that the server is started, check whether the set / etc/exports configuration is loaded correctly on the server.
$ showmount -e localhost Export list for localhost: /data/share 10.222.77.0/24
5. NFS test
Finally, test whether it can be mounted correctly on another Linux virtual machine. First, we can view the shared directory information of the NFS server (the IP of the upper server is 10.222.77.86) on the client.
# showmount -e 10.222.77.86 Export list for 10.222.77.86: /data/share 10.222.77.0/24
Then, create the hanging directory / share on the client
$ mkdir -p /share
Finally, mount the remote directory to the local / share directory.
$ mount 10.222.77.86:/data/share /share $ df -h | grep 10.222.77.86 Filesystem Size Used Avail Use% Mounted on 10.222.77.86:/data/share 27G 11G 17G 40% /share
As you can see, the remote NFS directory can be mounted locally correctly. Note: the mount point / share directory must already exist and there are no files or subdirectories in the directory.
Finally, we create a file in the / data/share directory of the NFS server to see if the client can read and modify it correctly.
# Server write $ echo "This is NFS server." > /data/share/nfs.txt # ll /data/share/ total 4 -rw-r--r-- 1 root root 20 Nov 5 16:49 nfs.txt # Client read $ ll /share/ total 4 -rw-r--r-- 1 root root 20 Nov 5 16:49 nfs.txt $ cat /share/nfs.txt This is NFS server. # Client write $ echo "This is NFS client." >> /share/nfs.txt # Server read $ cat /data/share/nfs.txt This is NFS server. This is NFS client.
This is because the NFS remote directory permission is set to rw and has read-write permission. If it is set to ro, the client can only read and cannot write. It is reasonably configured according to the actual application scenario, so it will not be demonstrated here. It is mentioned here that NFS is mounted using UDP protocol by default. In order to improve the stability of NFS, TCP protocol can be used for mounting. The following commands can be used for client mounting commands:
$ mount 10.222.77.86:/data/share /share -o proto=tcp -o nolock
Finally, if the client wants to uninstall NFS mount, use the following command.
$ umount /share
Well, the above briefly introduces the installation, configuration and use of NFS. Using it, we can easily share storage with different hosts and operating systems through the network. Next, we continue to study various ways to mount NFS in Kubernetes cluster to mount network file storage system.