Monitoring capacity through GUI

You can monitor the capacity of the file system, pools, filesets, NSDs, users, and user groups in the IBM Spectrum Scale system.

The capacity details displayed in the GUI are obtained from the following sources:

Based on the source of the capacity information, different procedures need to be performed to enable capacity and quota data collection.

For both GPFS quota database and performance monitoring tool-based capacity and quota collection, you need to use the Files > Quotas page to enable quota data collection per file system and enforce quota limit checking. If quota is not enabled for a file system:

To enable capacity data collection from the performance monitoring tool, the GPFSFilesetQuota sensor must be enabled. For more information on how to enable the performance monitoring sensor for capacity data collection, see Manual installation of IBM Spectrum Scale GUI in IBM Spectrum Scale: Concepts, Planning, and Installation Guide.

Capacity data obtained from the GPFS quota database

The capacity and quota information collected from the GPFS quota database is displayed on the Files > Quotas and Files > User Capacity pages in the management GUI.

1. Files > Quotas page

Use quotas to control the allocation of files and data blocks in a file system. You can create default, user, group, and fileset quotas through the Quotas page.

A quota is the amount of disk space and the amount of metadata that is assigned as upper limits for a specified user, group of users, or fileset. Use the Actions menu to create or modify quotas. The management GUI allows you to only manage capacity-related quota. The inode-related quota management is only possible in the command-line interface.

You can specify a soft limit, a hard limit, or both. When you set a soft limit quota, a warning is sent to the administrator when the file system is close to reaching its storage limit. A grace period starts when the soft quota limit is reached. Data is written until the grace period expires, or until the hard quota limit is reached. Grace time resets when used capacity goes below the soft limit. If you set a hard limit quota, you cannot save data after the quota is reached. If the quota is exceeded, you must delete files or raise the quota limit to store more data.
Note:
  • User or user group quotas for filesets are only supported if the Per Fileset option is enabled at the file system level. Use the command-line interface to set the option. See the manpages of mmcrfs and mmchfs commands for more detail.
  • You need to unmount a file system to change the quota enablement method from per file system to per fileset or vice versa.

You can set default user quotas at the file system level rather than defining user quotas explicitly for each user. Default quota limits can be set for users. You can specify the general quota collection scope such as per file system or per fileset to define whether the default quota needs to be defined at file system level or fileset level and set the default user quota. After this value is set, all child objects that are created under the file system or fileset will be configured with the default soft and hard limits. You can assign a custom quota limit to individual child objects, but the default limits remain the same unless changed at the file system or fileset level.

After reconfiguring quota settings, it is recommended to run the mmcheckquota command for the affected file system to verify the changes.

For more information on how to manage quotas, see Managing GPFS quotas section in the IBM Spectrum Scale: Administration Guide.

Capacity data from users, groups, and filesets with no quota limit set are not listed in the Quotas page. Use the Files > User Capacity page to see capacity information of such users and groups. Use the Files > Filesets page to view current and historic capacity information of filesets.

2. Files > User Capacity page

The Files > User Capacity page provides predefined capacity reports for users and groups. While capacity information of file systems, pools, and filesets is available in the respective areas of the GUI, the Files > User Capacity page is the only place where information on used capacity per user or group is available.

The User Capacity page depends on the quota accounting method of the file system. You need to enable quota for a file system to display the user capacity data. If quota is not enabled, you can follow the fix procedure in the Files > Quotas page or use the mmchfs <Device> -Q yes CLI command to enable quota. Even if the capacity limits are not set, the User Capacity page shows data as soon as the quota accounting is enabled and users write data. This is different in the Quotas page, where only users and groups with quota limits defined are listed. The user and group capacity quota information is automatically collected once a day by the GUI.

For users and user groups, you can see the total capacity and whether quotas are set for these objects. you can also see the percentage of soft limit and hard limit usage. When the hard limit is exceeded, no more files belong to the respective user, user group, or fileset can be written. However, exceeding the hard limit allows a certain grace period before disallowing more file writes. Soft and hard limits for disk capacity are measured in units of kilobytes (KiB), megabytes (MiB), or gigabytes (GiB). Use the Files > Quotas page to change the quota limits.

Capacity data collected through the performance monitoring tool

The historical capacity data collection for file systems, pools, and filesets depend on the correctly configured data collection sensors for fileset quota and disk capacity. When the IBM Spectrum Scale system is installed through the installation toolkit, the capacity data collection is configured by default. In other cases, you need to enable capacity sensors manually.

If the capacity data collection is not configured correctly you can use mmperfmon CLI command or the Services > Performance Monitoring > Sensors page.

The Services > Performance Monitoring > Sensors page allows to view and edit the sensor settings. By default, the collection periods of capacity collection sensors are set to collect data with a period of up to one day. Therefore, it might take a while until the data is refreshed in the GUI.

The following sensors are collecting capacity related information and are used by the GUI.
GPFSDiskCap
NSD, Pool and File system level capacity. Uses the mmdf command in the background and typically runs once per day as it is resource intensive. Should be restricted to run on a single node only.
GPFSPool
Pool and file system level capacity. Requires a mounted file system and typically runs every 5 minutes. Should be restricted to run on a single node only.
GPFSFilesetQuota
Fileset capacity based on the quota collection mechanism. Typically, runs every hour. Should be restricted to run only on a single node.
GPFSFileset
Inode space (independent filest) capacity and limits. Typically runs every 5 minutes. Should be restricted to run only on a single node.
DiskFree
Overall capacity and local node capacity. Can run on every node.

The Monitoring > Statistics page allows to create customized capacity reports for file systems, pools and filesets. You can store these reports as favorites and add them to the dashboard as well.

The dedicated GUI pages combine information about configuration, health, performance, and capacity in one place. The following GUI pages provide the corresponding capacity views:

  • Files > File Systems
  • Files > Filesets
  • Storage > Pools
  • Storage > NSDs

The Filesets grid and details depend on quota that is obtained from the GPFS quota database and the performance monitoring sensor GPFSFilesetQuota. If quota is disabled, the system displays a warning dialog in the Filesets page.

Troubleshooting issues with capacity data displayed in the GUI

Due to the impact that capacity data collection can have on the system, different capacity values are collected on a different schedule and are provided by different system components. The following list provides insight on the issues that can arise from the multitude of schedules and subsystems that provide capacity data:

Capacity in the file system view and the total amount of the capacity for pools and volumes view do not match.
The capacity data in the file system view is collected every 10 minutes by performance monitoring collector, but the capacity data for pools and Network Shared Disks (NSD) are not updated. By default, NSD data is only collected once per day by performance monitoring collector and it is cached. Clicking the refresh icon gathers the last two records from performance monitoring tool and it displays the last record values if they are not null. If the last record has null values, the system displays the previous one. If the values of both records are null, the system displays N/A and the check box for displaying a time chart is disabled. The last update date is the record date that is fetched from performance monitoring tool if the values are not null.
Capacity in the file system view and the total amount of used capacity for all filesets in that file system do not match.
There are differences both in the collection schedule as well as in the collection mechanism that contributes to the fact that the fileset capacities do not add up to the file system used capacity.

Scheduling differences:

Capacity information that is shown for filesets in the GUI is collected once per hour by performance monitoring collector and displayed on Filesets page. When you click the refresh icon you get the information of the last record from performance monitoring. If the last two records have null values, you get a 'Not collected' warning for used capacity. The file system capacity information on the file systems view is collected every 10 minutes by performance monitoring collector and when you click the refresh icon you get the information of the last record from performance monitoring.

Data collection differences:

Quota values show the sum of the size of all files and are reported asynchronously. The quota reporting does not consider metadata, snapshots, or capacity that cannot be allocated within a subblock. Therefore, the sum of the fileset quota values can be lower than the data shown in the file system view. You can use the CLI command mmlsfileset with the -d and -i options to view capacity information. The GUI does not provide a means to display this values because of the performance impact due to data collection.

The sum of all fileset inode values on the view quota window does not match the number of inodes that are displayed on the file system properties window.
The quota value only accounts for user-created inodes while the properties for the file system also display inodes that are used internally. Refresh the quota data to update these values.
No capacity data shown on a new system or for a newly created file system
Capacity data may show up with a delay of up to 1 day. The capacity data for file systems, NSDs, and pools is collected once a day as this is a resource intensive operation. Line charts do not show a line if only a single data point exists. You can use the hover function in order to see the first data point in the chart.
The management GUI displays negative fileset capacity or an extremely high used capacity like millions of Petabytes or 4000000000 used inodes.
This problem can be seen in the quota and filesets views. This problem is caused when the quota accounting is out of sync. To fix this error, issue the mmcheckquota command. This command recounts inode and capacity usage in a file system by user, user group, and fileset, and writes the collected data into the database. It also checks quota limits for users, user groups, and filesets in a file system. Running this command can impact performance of I/O operations.
No capacity data is displayed on the performance monitoring charts
Verify whether the GPFSFilesetQuota sensor is enabled. You can check the sensor status form the Services > Performance Monitoring page in the GUI. For more information on how to enable the performance monitoring sensor for capacity data collection, see Manual installation of IBM Spectrum Scale GUI in IBM Spectrum Scale: Concepts, Planning, and Installation Guide.