Previous topic

find

Next topic

fsck

This Page

fs

fs add|boot|config|dropdeletion|dropghosts|dropfiles|dumpmd|ls|mv|rm|status [OPTIONS]
  Options:
  fs add [-m|--manual <fsid>] <uuid> <node-queue>|<host>[:<port>] <mountpoint> [<space>] [<status>]
  add and assign a filesystem based on the unique identifier of the disk <uuid>
  -m|--manual  : add with user specified <fsid> and <space>
  <fsid>       : numeric filesystem id 1...65535
  <uuid>       : unique string identifying current filesystem
  <node-queue> : internal EOS identifier for a node e.g /eos/<host>:<port>/fst
  it is preferable to use the host:port syntax
  <host>       : FQDN of host where filesystem is mounter
  <port>       : FST XRootD port number [usually 1095]
  <mountponit> : local path of the mounted filesystem e.g /data/
  <space>      : space in which to insert the filesystem, if nothing is
  specified then space "default" is used
  <status>     : set filesystem status after insertion e.g off|rw|ro etc.
fs boot <fsid>|<uuid>|<node-queue>|* [--syncmgm]
boot - filesystem identified by <fsid> or <uuid>
- all filesystems on a node identified by <node-queue>
- all filesystems registered
--syncmgm    : for MGM resynchronization during the booting
fs clone <sourceid> <targetid>
replicate files from the source to the target filesystem
<sourceid>   : id of the source filesystem
<targetid>   : id of the target filesystem

fs compare <sourceid> <targetid>
compares and reports which files are present on one filesystem and not on the other
<sourceid>   : id of the source filesystem
<targetid>   : id of the target filesystem

fs config <fsid> <key>=<value>
configure the filesystem parameter, where <key> and <value> can be:
configstatus=rw|wo|ro|drain|draindead|off|empty [--comment "<comment>"]
rw        : set filesystem in read-write mode
wo        : set filesystem in write-only mode
ro        : set filesystem in read-only mode
drain     : set filesystem in drain mode
draindead : set filesystem in draindead mode, unusable for any read
off       : disable filesystem
empty     : empty filesystem, possible only if there are no
more files stored on it
--comment : pass a reason for the status change
headroom=<size>
headroom to keep per filesystem. <size> can be (>0)[BMGT]
scaninterval=<seconds>
entry rescan interval (default 7 days), 0 disables scanning
scanrate=<MB/s>
maximum IO scan rate per filesystem
scan_disk_interval=<seconds>
disk consistency thread scan interval (default 4h)
scan_ns_interval=<seconds>
namespace consistency thread scan interval (default 3 days)
scan_ns_rate=<entries/s>
maximum scan rate of ns entries for the NS consistency. This
is bound by the maxium number of IOPS per disk.
graceperiod=<seconds>
grace period before a filesystem with an operation error gets
automatically drained
drainperiod=<seconds>
period a drain job is allowed to finish the drain procedure
proxygroup=<proxy_grp_name>
schedule a proxy for the current filesystem by taking it from
the given proxy group. The special value "<none>" is the
same as no value and means no proxy scheduling
filestickyproxydepth=<depth>
depth of the subtree to be considered for file-stickyness. A
negative value means no file-stickyness
forcegeotag=<geotag>
set the filesystem's geotag, overriding the host geotag value.
The special value "<none>" is the same as no value and means
no override
s3credentials=<accesskey>:<secretkey>
the access and secret key pair used to authenticate
with the S3 storage endpoint
fs dropdeletion <fsid>
drop all pending deletions on the filesystem
fs dropghosts <fsid> [--fxid fid1 [fid2] ...]
drop file ids (hex) without a corresponding metadata object in
the namespace that are still accounted in the file system view.
If no fxid is provided then all fids on the file system are checked.
fs dropfiles <fsid> [-f]
drop all files on the filesystem
-f : unlink/remove files from the namespace (you have to remove
the files from disk)
fs dumpmd <fsid> [--fid] [--path] [--size] [-m|-s]
dump all file metadata on this filesystem in query format
--fid  : dump only the file ids
--path : dump only the file paths
--size : dump only the file sizes
-m     : print full metadata record in env format
-s     : silent mode (will keep an internal reference)
fs ls [-m|-l|-e|--io|--fsck|[-d|--drain]|-D|-F] [-s] [-b|--brief] [[matchlist]]
list filesystems using the default output format
-m         : monitoring format
-b|--brief : display hostnames without domain names
-l         : display parameters in long format
-e         : display filesystems in error state
--io       : IO output format
--fsck     : display filesystem check statistics
-d|--drain : display filesystems in drain or draindead status
along with drain progress and statistics
-D|--drain_jobs :
display ongoing drain transfers, matchlist needs to be an integer
representing the drain file system id
-F|--failed_drain_jobs :
display failed drain transfers, matchlist needs to be an integer
representing the drain file system id. This will only display
information while the draining is ongoing
-s         : silent mode
[matchlist]
-> can be the name of a space or a comma separated list of
spaces e.g 'default,spare'
-> can be a grep style list to filter certain filesystems
e.g. 'fs ls -d drain,bootfailure'
-> can be a combination of space filter and grep e.g.
'fs ls -l default,drain,bootfailure'
fs mv [--force] <src_fsid|src_grp|src_space> <dst_grp|dst_space>
move filesystem(s) in different scheduling group or space
--force   : force mode - allows to move non-empty filesystems bypassing group
and node constraints
src_fsid  : source filesystem id
src_grp   : all filesystems from scheduling group are moved
src_space : all filesystems from space are moved
dst_grp   : destination scheduling group
dst_space : destination space - best match scheduling group
is auto-selected
fs rm <fsid>|<mnt>|<node-queue> <mnt>|<hostname> <mnt>
remove filesystem by various identifiers, where <mnt> is the
mountpoint
fs status [-r] [-l] <identifier>
return all status variables of a filesystem and calculates
the risk of data loss if this filesystem is removed
<identifier> can be:
<fsid> : filesystem id
[<host>] <mountpoint> : if host is not specified then it's
considered localhost
-l : list all files which are at risk and offline files
-r : show risk analysis
Examples:
fs ls --io -> list all filesystems with IO statistics
fs boot *  -> send boot request to all filesystems
fs dumpmd 100 -path -> dump all logical path names on filesystem 100
fs mv 100 default.0 -> move filesystem 100 to scheduling group default.0