Skip to content

Latest commit

 

History

History
113 lines (79 loc) · 3.53 KB

cli_overview.md

File metadata and controls

113 lines (79 loc) · 3.53 KB
Raw

Growing LCFS file system

When the underlying device/file is extended in size, the LCFS can make use of the additional space after running the following command.

# sudo lcfs grow /lcfs

Adjusting the frequency of LCFS commit (sync) operations

Periodically, LCFS commits its state in memory to disk to make those persistent. All operations happened until that point will be saved on disk and will not be lost even if an abnormal shutdown happens (system crashes and such). This background activity is triggered every minute by default. This time interval could be changed by running the following command.

# sudo lcfs syncer /lcfs <time in seconds>

This background activity can be disabling by specifying 0 seconds.

Adjusting the amount of memory for caching images in memory

LCFS caches images in memory in a private page cache. It is configured to use 512MB or 5% of the system memory (whichever is higher) by default. This limit could be changed by running the following command.

# sudo lcfs pcache /lcfs <memory limit in MB>

Releasing memory used to cache images

The memory used for caching images (private page cache) could be freed by running the following command.

# sudo lcfs flush /lcfs

Trigger a commit (sync) operation

If needed, all dirty data in memory could be committed to disk by running the following command.

# sudo lcfs commit /lcfs

Options which can be enabled at mount time

A few capabilities of LCFS are not turned on by default for performance reasons. Those could be enabled by specifying appropriate options while mounting the LCFS. Here is a complete list of options.

usage: lcfs daemon <device/file> <host-mountpath> <plugin-mountpath> [-f] [-c] [-d] [-m] [-r] [-t] [-p] [-s] [-v]
    device     - device or file - image layers will be saved here
    host-mount - mount point on host
    host-mount - mount point propogated the plugin
    -f         - run foreground (optional)
    -c         - format file system (optional)
    -d         - display fuse debugging info (optional)
    -m         - enable memory stats (optional)
    -r         - enable request stats (optional)
    -t         - enable tracking count of file types (optional)
    -p         - enable profiling (optional)
    -s         - swap layers when committed
    -v         - enable verbose mode (optional)

Stats

Various stats could be displayed by running the following command.

# sudo lcfs stats /lcfs <layer id or .> [-c]

The stats will be logged into syslog. Stats will be displayed for the layer specified (layer id), or for all the layers if . is specified as the layer id.

Stats could be cleared before running some experiments by specifying -c option with the above command.

Stats are not collected by default for performance reasons. Different types of stats need to be enabled while mounting the LCFS by specifying the appropriate options. Here is a list of stats supported as of now.

Memory stats

Each malloc() and free() calls made by LCFS is tracked when memory stats are enabled. This is enabled by specifying -m option.

Request stats

All file operations and ioctl requests are counted and times taken for each of them are tracked for each layer separately when request stats are enabled by specifying -r option.

File types

Type of files (regular, directories, symbolic links, other) created in every

Profiling

If profiling is enabled at mount time, it will be saved under /tmp/lcfs when LCFS is unmounted. The saved profiles could be examined using gperftools commands like pprof.