Resolve the issue of insufficient disk space on a Linux instance - Elastic Compute Service

Typically, your Linux instance has a limited number of disk space. If your application is deployed on a Linux instance, the available disk space decreases as the application continues to run and the number of stored files increases. When you have insufficient disk space, the service cannot continue to write files to the disk, which may cause service exceptions. This topic describes the method of checking for the insufficient disk space issue and how to resolve the issue.

Problem description

If the No space left on device error message is returned when you create a file or application on a Linux Elastic Compute Service (ECS) instance, the disk space on the instance is insufficient. This topic describes the causes of the issue of insufficient disk space and how to resolve the issue.

Possible causes

In most cases, the issue of insufficient disk space occurs due to the following causes:

The space usage of disk partitions reaches 100%.
The inode usage of disk partitions reaches 100%.
Leftover files (zombies) exist.
Note
A deleted file may fail to free up file space when the file is deleted because its file handle was open before it was deleted.
Mount points are overwritten.
Note
For example, the file system of a disk partition that manages a large number of files is mounted on a directory. The directory is considered a mount point. If the file system of another disk partition is mounted on the mount point, the mount point is overwritten by the new file system. However, your application may continue to read data from and write data to the original file system space. In this case, an error message that indicates the issue of insufficient disk space may be returned. If you run the df or du command to check the disk space usage of files and directories, no information is returned. This is because the df or du command returns information about the space usage of the disk partition that corresponds to the current mount point.
The upper limit on inotify watches is reached.
Note
The inotify API provides a mechanism for monitoring file system events in Linux. You can use inotify to monitor file changes in file systems in real time. This error is not caused by insufficient disk space. This topic describes the error to help you troubleshoot issues.

Troubleshooting methods and solutions

To resolve the No space left on device error, perform the following operations based on the cause of the issue:

Note

You must connect to the ECS instance on which you can troubleshoot the issue. For more information, see Use Workbench to log on to a Linux instance over SSH.

Warning

In the following operations, if you want to release disk space by deleting a file, make sure that the file is no longer needed before you delete it. This prevents data loss or impacts on your business. We recommend that you perform a disk backup before you manually delete the files. You can copy the files or use snapshots to back up data. For more information about how to create a snapshot, see Create a snapshot for a disk.

The space usage of disk partitions reaches 100%

Identification methods

Query the current usage of disk space and find files that consume a large amount of disk space.

Run the following command to query the usage of disk space:
```
df -h
```
The following command output is returned: In the example, the usage of the /dev/vda3 partition reaches 100%.
Run the following commands to go to the root directory and identify which directory consumes the largest amount of disk space:
```
sudo du -sh /* | sort -rh | head -n 10
```
The following command output is returned. As shown in the figure, the /home directory consumes the largest space. Therefore, check which file or directory in the /home directory occupies the largest space. Perform the following operations based on actual scenarios.
Run the following command to identify which directory consumes the largest amount of disk space.
For example, the preceding command output indicates that the /home directory consumes the largest amount of disk space. Then, you need to further identify which file or directory consumes the largest amount of disk space in the /home directory.
```
sudo du -sh /home/* | sort -rh | head -n 10
```
The following command output is returned. The ecs-user directory consumes the largest space. You can further check the specific files or subdirectories under the ecs-user directory to determine which items consumes the larger space.
Check whether the large files in the /home/ecs-user/ directory can be deleted.

Solutions

Perform one of the following operations based on the cause and your business scenario:

Delete files or directories that consume a large amount of disk space and are no longer needed based on your business requirements.
Resize existing disks or create more disks based on your business requirements if your business does not allow you to delete files from disks or disk space cannot be released by deleting files. For more information, see Overview of disk resizing, Create an empty data disk, and Attach a data disk.

The inode usage of disk partitions reaches 100%

Each file or directory in a file system is identified by a unique inode. A specific number of inodes are pre-allocated to a disk partition for files and directories when the disk partition is formatted. However, if the file system of the disk partition stores a large number of small files or directories, inode resources may become insufficient. If all inodes of a disk partition are allocated to files or directories, no new files or directories can be created on the disk partition regardless of whether the disk partition has free space. To resolve the issue, you can delete unnecessary files or directories to release inodes or increase the number of inodes.

Identification method

Run the following command to query the inode usage:

df -i

Solutions

If the inode usage reaches or is about to reach 100%, use one of the following methods to resolve the issue.

Clean up files or directories that are heavily consumed by inodes to reduce inode usage
If you do not want to format disks to increase the number of inodes, perform the following operations to delete files or directories that use a large number of inodes:
1. Run the following command to query the number of files that exist in each subdirectory of the root directory:
```
for i in /*; do echo $i; sudo find $i | wc -l; done
```
  The sample command output shown in the following figure is displayed. The command output indicates that the /mnt directory has the largest number of files. Then, you need to identify which directory in the /mnt directory has the largest number of files. Inode usage increases as the number of files increases. Perform the following operations based on actual scenarios:
2. Run the following command to see which file or directory in the /mnt directory with the highest inode usage has the highest inode usage.
3. You can then locate the file or directory with the largest inode usage and clean up the space.
Reformat the disk to increase the number of inodes
If files stored on disks cannot be deleted or if the inode usage remains high after files are deleted, perform the following steps to increase the number of inodes in the file systems: Back up disk data, format the disks, and then copy the data back to the disks.
1. Run the following command to query the disk partition format:
```
lsblk -f
```
  The following sample command output is returned.
  The partition format of the destination disk is Ext4.
2. The following example shows how to increase the number of inode based on your actual disk partition format.
  Ext* file system
  Warning
  You must unmount the file systems from mount points. However, this may interrupt your application services. We recommend that you unmount the file systems during an appropriate period of time.
  The disks must be formatted. In this case, the data stored on the disks is deleted. Before you format the disks, you must back up the data. You can copy files or create snapshots to back up data. For more information about how to create a snapshot, see Create a snapshot.
  Run the following command to unmount the file system: In this example, the /mnt/device_vdc file system is uninstalled.
  sudo umount /mnt/device_vdc
  Run the following commands to rebuild the file system and increase the number of inode nodes:
  In this example, an Ext4 file system is created for the /dev/vdc partition and the number of inodes is set to 163,840. Modify the command based on the actual scenario.
  sudo mkfs.ext4 /dev/vdc -N 163840
  Note
  In Linux, the number of inodes varies based on the disk capacity. In most cases, the number of inodes is calculated by using the following formula: Number of inodes = Disk capacity (KB)/16 KB. For example, a 40 GB disk can have 2,621,440 inodes based on the preceding formula. The maximum number of inodes allowed for a disk is 2^32, which is approximately 4.3 billion. Specify the number of inodes based on the disk capacity.
  Run the following command to mount the new file system to the mount point. In this example, the /dev/vdc partition is mounted to the /mnt/device_vdc/ directory.
  sudo mount /dev/vdc1 /mnt/device_vdc/
  (Optional) Run the following command to query the number of inodes and check whether the number of inodes is increased:
  df -i
  The following command output indicates that the inode quantity is changed as expected. You can continue to copy the backup data to restore related data or applications.
  XFS file system
  Warning
  To increase the number of inodes, you must unmount the file systems from mount points. However, this may interrupt your application services. We recommend that you unmount the file systems during an appropriate period of time.
  To increase the number of inodes on disks, the disks must be formatted. In this case, the data stored on the disks is deleted. Before you format the disks, you must back up the data. You can copy files or create snapshots to back up data. For information about how to create a snapshot, see Create a snapshot.
  Run the following command to unmount the file system: In this example, the /mnt/device_vdc is uninstalled.
  sudo umount /mnt/device_vdc
  Run the following commands to re-establish the file system and increase the number of Inode nodes:
  In this example, the disk partition is /dev/vdc, the file system type is xfs, and the default maxpct=25 is changed to 40.
  sudo mkfs.xfs -f -i maxpct=40 /dev/vdc
  Note
  In Linux, the number of inodes in the XFS partition format varies based on the disk capacity and is affected by the disk capacity and the maxpct parameter. By default, the proportion of inodes in a file system with a capacity of less than 1 TB is 25%, the proportion of inodes in a file system with a capacity of less than 50 TB is 5%, and the proportion of inodes in a file system with a capacity of more than 50 TB is 1%. You can specify the number of inodes that suits your business needs.
  Run the following command to mount the new file system to the mount point. In this example, the /dev/vdc device is mounted to the /mnt/device_vdc/ directory.
  sudo mount /dev/vdc /mnt/device_vdc/
  (Optional) Run the following command to query the number of inodes and check whether the number of inodes is increased:
  df -i
  The sample command output shown in the following figure indicates that the number of inodes is changed. You can copy backup data back to the disk and restore the affected applications.

Leftover files (zombies) exist

If the disk partition capacity and inode capacity are both normal, a large number of deleted files (deleted) may exist in the system. These files are still occupied by processes in the system, causing the system to fail to release the corresponding disk space. These files have been marked as deleted and cannot be counted by running the df or du command. An excessively large number of leftover files occupy a large amount of disk space. Perform the following operations to query and delete leftover files:

Identification methods

If lsof is not installed on the operating system, run one of the following commands to install lsof.
Alibaba Cloud Linux and CentOS
```
sudo yum install -y lsof
```
Debian and Ubuntu
```
sudo apt install -y lsof
```
Run the following command to query the disk space usage of leftover files:
```
sudo lsof | grep delete | sort -k7 -rn | more
```
The sample command output shown in the following figure is displayed. The sizes of files that are in the deleted state are displayed in the seventh column. Check whether the sum of the sizes is close to the unexpected usage of disk space. If the sum is close to the unexpected usage, leftover files consume the disk space.

Solutions

If leftover files exist, use one of the following methods to release the file handles and clear the leftover files to release the disk space:

Restart the instance
After the instance restarts, the system terminates running processes and releases the handles of deleted files.
Warning
Instance restarts may interrupt services. We recommend that you restart instances during an appropriate period of time.
Run the kill command
After you run the lsof command, the process IDs (PIDs) of running processes that correspond to the leftover files are usually displayed in the second column. You can specify a PID in the kill command to terminate the corresponding process.
1. Run the following command to list the PIDs of processes:
```
lsof | grep delete 
```
2. Run the following command to terminate a process that corresponds to a leftover file based on your business requirements:
```
kill <PID>
```
  Warning
  When processes are terminated, services that run on instances may be affected. Proceed with caution.

Mount points are overwritten

If you cannot determine the cause of insufficient disk space after the preceding three causes are eliminated, the mount point may be overwritten. You can use the following methods to find the cause.

Identification methods

Run the following command to view the mount information of the data disk:
```
mount
```
If a command output similar to the following one is returned, the JDK is installed:
Two devices are mounted to the /mnt/device_vdc directory. Mount points may be overwritten in the directory.
Run the following command to view information about the partitions:
```
df -h
```
The following command output is returned.

Solutions

To resolve this issue, unmount the disk partition and check the space usage of the original mount point.

Run the following command to unmount the file system: In this example, the /mnt/device_vdc file system is uninstalled.

Warning

To increase the number of inodes, you must unmount the file systems from mount points. However, this may interrupt your application services. We recommend that you unmount the file systems during an appropriate period of time.

sudo umount /mnt/device_vdc

After you cancel the mounting of a disk partition, you must check the space usage of the original mount directory and take appropriate measures based on the specific situation.

The upper limit on inotify watches is reached

If an error message similar to tail: cannot watch '...': No space left on device is returned when you run commands such as tail -f, the upper limit on inotify watches is reached. To resolve the issue, you can increase the upper limit on inotify watches.

Identification methods

Run the following command to view the upper limit on inotify watches:

cat /proc/sys/fs/inotify/max_user_watches

Solutions

Run the following command to change the upper limit on inotify watches:

sudo sysctl fs.inotify.max_user_watches=<New upper limit>

Replace <New upper limit> with the value of the upper limit that you want to specify for inotify watches.

Note

If you increase the upper limit on inotify watches, inotify watches may occupy a larger amount of system memory. When you change the upper limit on inotify watches, evaluate the memory and performance of the system and possible impact. You can run the man 7 inotify command to learn more about inotify watches and the related settings.

References

If your disk space is often insufficient or your data storage requirements increase, we recommend that you optimize the storage solution based on your business requirements. For more information, see the following optimization suggestions.

If your disk space stores a large number of files, such as images and videos, and no high-concurrency read or write operations are performed on the disks, you can use Object Storage Service (OSS). OSS is a large-capacity, secure, cost-effective, and highly reliable cloud storage service that can automatically expand storage space based on the volume of data. You can use ossfs to mount OSS buckets to ECS instances. Applications can manage objects in OSS buckets in the same manner as on-premises files without the need to modify code. For information about how to use ossfs to mount an OSS bucket to a local directory on a Linux system, see ossfs.
If your business requires high-concurrency read and write operations and data sharing, you can use Apsara File Storage NAS (NAS) to store files. NAS provides simple and scalable file systems that allow high-performance and high-concurrency shared storage, and can be used together with ECS. NAS can automatically expand the storage space as the data volume increases without the need for manual operations. For more information, see Mount a file system on a Linux ECS instance.
If you store a large number of log files on disks, you can transfer the log files to Simple Log Service. This facilitates log query and reduces disk usage. For more information, see Getting Started.

Problem description

Possible causes

Troubleshooting methods and solutions

The space usage of disk partitions reaches 100%

Identification methods

Solutions

The inode usage of disk partitions reaches 100%

Identification method

Solutions

Ext* file system

XFS file system

Leftover files (zombies) exist

Identification methods

Alibaba Cloud Linux and CentOS

Debian and Ubuntu

Solutions

Mount points are overwritten

Identification methods

Solutions

The upper limit on inotify watches is reached

Identification methods

Solutions

References