| title | expires_at | tags | |
|---|---|---|---|
Overview |
never |
|
Go-based FUSE file system for uid mapping local file access
In CloudFoundry NFS volume services,
the NFS driver will start mapfs when the uid or gid parameters
are specified when binding an NFS mount. MapFS will expose a FUSE filesystem that is mounted into the app container,
and the app can read and write from. This is backed by an NFS mount.
MapFS replicates file access so that as far as NFS is concerned, the file access is done
with the uid and gid specified.
Because an app can potentially go faster than NFS (which is limited by network bandwidth), it's possible for requests to stack up in MapFS while it waits for NFS to respond. Currently, go-fuse does not implement backpressure or load shedding. For every request in-flight, a buffer and a goroutine will be created, so MapFS can consume high memory and processor when the app accesses the filesystem faster than NFS responds.
There is an optional configuration file that can be placed at /var/vcap/jobs/map-fs/config/mapfs.yml that will allow
MapFS to be configured. Configuration keys are:
debug- enables go-fuse debug mode. Logs are written to/var/vcap/sys/log/map-fs/mapfs.$$.logwhere$$is the process ID.single_threaded- enables go-fuse single threaded mode. Can reduce memory pressure, and may slow down access.cpu_profile- writes a Go pprof CPU profile to the specified path.mem_profile- when the process receives aSIGUSR1a Go pprof memory profile will be written to the path with a timestamp as a file suffix.soft_mem_limit- soft memory limit in bytes. Equivalent to settingGOMEMLIMITenvironment variable.
The configuration file is read on startup. To restart MapFS just kill -9 the existing
process and the NFS Driver will start a new one. This may break running apps.
map-fs job is added to cf-deployment when using enable-nfs-volume-service.yml