This topic describes the limits of NAS on product specifications, operating systems, protocols, and the classic network.
Limits on product specifications
|Maximum number of file systems under each Alibaba Cloud account in a region||20|
|Maximum number of mount targets for each file system||
|Maximum number of files in each file system||1 billion|
|Maximum directory depth||1,000 levels|
|Maximum capacity of each file system||
|Maximum number of compute nodes on which a file system is mounted||1,000|
|Maximum size of each file||32 TB|
|Maximum number of hard links to each file||511|
Limits on operating systems
- Alibaba Cloud has tested the official kernel images and listed the recommended images. If you do not use the recommended kernel images, Alibaba Cloud does not guarantee the availability of the NAS service. For more information about the recommended kernel versions, see Recommended kernel images.
- Alibaba Cloud has identified multiple defective kernel versions and provided solutions. If you use a defective kernel version and do not upgrade the version, Alibaba Cloud does not guarantee the availability of the NAS service. For more information about the defective kernel versions, see Known issues on an NFS client.
- If Alibaba Cloud identifies a new defective kernel version, Alibaba Cloud will notify you by using an email, internal message, SMS message, and console announcement. Make sure that you can receive the notifications and replace the defective kernel version at the earliest opportunity. Otherwise, Alibaba Cloud does not guarantee the availability of the NAS service.
Limits on protocols
- Network File System (NFS) file systems
- The maximum length of a file name is 255 bytes.
- NFSv4 does not support the following attributes: FATTR4_MIMETYPE, FATTR4_QUOTA_AVAIL_HARD, FATTR4_QUOTA_AVAIL_SOFT, FATTR4_QUOTA_USED, FATTR4_TIME_BACKUP, and FATTR4_TIME_CREATE. If you use an unsupported attribute, an NFS4ERR_ATTRNOTSUPP error is logged in the /var/log/messages file.
- NFSv4.1 does not support the following attributes: FATTR4_DIR_NOTIF_DELAY, FATTR4_DIRENT_NOTIF_DELAY, FATTR4_DACL, FATTR4_SACL, FATTR4_CHANGE_POLICY, FATTR4_FS_STATUS, FATTR4_LAYOUT_HINT, FATTR4_LAYOUT_TYPES, FATTR4_LAYOUT_ALIGNMENT, FATTR4_FS_LOCATIONS_INFO, FATTR4_MDSTHRESHOLD, FATTR4_RETENTION_GET, FATTR4_RETENTION_SET, FATTR4_RETENTEVT_GET, FATTR4_RETENTEVT_SET, FATTR4_RETENTION_HOLD, FATTR4_MODE_SET_MASKED, and FATTR4_FS_CHARSET_CAP. If you use an unsupported attribute, an NFS4ERR_ATTRNOTSUPP error is logged in the /var/log/messages file.
- NFSv4 does not support the following operations: OP_DELEGPURGE, OP_DELEGRETURN, and NFS4_OP_OPENATTR. If you call an unsupported operation, an NFS4ERR_ATTRNOTSUPP error is logged in the /var/log/messages file.
- NFSv4 does not support the delegation feature.
- UID and GID
- In Linux, mappings between user IDs (UIDs) and usernames and mappings between group IDs (GIDs) and group names are defined in configuration files. For NFSv3 file systems, if the mapping between an ID and a name is defined in a configuration file, the name is used to represent the user or group. If no mapping is defined for an ID, the ID is used.
- For NFSv4 file systems, the usernames and group names of all files are identified
as nobody if the version of the Linux kernel is earlier than 3.0. If the kernel version
is later than 3.0, the rule that is used by NFSv3 file systems applies.
Note If the Linux kernel version is earlier than 3.0, we recommend that you do not run the chown or chgrp command for files or directories that are stored on a mounted NFSv4 file system. Otherwise, the UID and GID of the file or directory will be changed to nobody.
- Limits on NFS clients
- For General-purpose NAS file systems, an NFS client allows a maximum of 128 concurrent requests.
- For General-purpose NAS file systems, the maximum bandwidth of an NFS client is 500 MB/s.
- You can open a maximum of 32,768 files at a time on an NFS client. Files in the list folder and its subfolders are not counted.
- Each mount on an NFS client can acquire up to 8,192 locks across a maximum of 256 file-process pairs. For example, a single process can acquire one or more locks on 256 separate files, or 8 processes can each acquire one or more locks on 32 files.
- We recommend that you do not use an NFS client in a Windows host to access an NFS file system.
- Server Message Block (SMB) file systems
- The maximum length of a file name is 255 UTF-16 characters.
- SMB file systems do not support extended file attributes and client-side caching based on leases.
- SMB file systems do not support I/O control (IOCTL) or file system control (FSCTL) operations, such as creating sparse files, compressing files, inspecting the network interface card (NIC) status, and creating reparse points.
- SMB file systems do not support alternate data streams.
- SMB file systems do not support identity authentication that is provided by Lightweight Directory Access Protocol (LDAP).
- SMB file systems do not support several features that are provided by SMB 3.0 or later, such as SMB Direct, SMB Multichannel, SMB Directory Leasing, and Persistent File Handle.
- Protocol versions
SMB 2.1 and later versions are supported. Compared with SMB 2.1 and later versions, SMB 1.0 provides lower performance and fewer features. Furthermore, Windows products that support only SMB 1.0 have reached end of support.
- Limits on SMB clients
Each file or directory can be opened a maximum of 8,192 times at a time. This limit applies across all compute nodes that have the file system mounted and all users that share access to the file system. This indicates a maximum of 8,192 active file handles for each file or directory. A maximum of 65,536 active file handles can exist on a file system.
Limits on the classic network
- If you mount a NAS file system on an ECS instance in the classic network, the file system and the ECS instance must belong to the same Alibaba Cloud account. If the ECS instance is transferred to another Alibaba Cloud account, the file system becomes inaccessible. In this case, Alibaba Cloud does not guarantee the availability of the NAS service.
- If you mount a NAS file system on an ECS instance in the classic network, the IP address of the mount target changes under specific circumstances. Due to the limits of the kernel, you must remount the NAS file system. If the IP address of a mount target in the classic network changes, Alibaba Cloud will notify you by using an email, internal message, SMS message, voice call, and console announcement 30 days in advance. If you do not remount the NAS file system after 30 days since the date of the notification, Alibaba Cloud cannot guarantee the availability of the NAS service.