Scalla

File Residency Manager Reference

Andrew Hanushevsky

1-May-2009

Produced by Andrew Hanushevsky for Stanford University under contract

DE-AC02-76-SFO0515 with the Department of Energy.

1 Introduction

This document describes File Residency Manager (FRM) commands and configuration directives. FRM was designed to enhanced support of the Open Storage System (oss) scalable file system, an xrootd extended feature and is meant to replace the deprecated MPS scripts. This feature is enabled when the Open File System (ofs) is loaded via libXrdOfs.so, at run-time using the xrootd.fslib configuration directive. FRM can be used to administer the local file system (i.e., placing files and maintaining symbolic links to these files) and for interfacing the file system to a Mass Storage System (MSS). The FRM components are fully configurable so that virtually any MSS and data transfer mechanisms can be used.

Using the oss component does not require that you use FRM. In fact, there is no reason to use FRM if you are not using a real or virtual MSS nor have enabled scalable file system support using the oss.cache configuration directive.

1.1 FRM Components

The FRM consists of four configurable components:

· Migration,

· Pre-staging,

· Purging,

· Data Transfer, and

· File management.

Each component is implemented as independent commands:

· frm_admin implements file administration,

· frm_migrd migrates modified files from local disk to a remote MSS,

· frm_pstga and frm_pstgd implement pre-staging,

· frm_purgd purges unreferenced files from local disk, and

· frm_xfr[.species] transfers data to/from the MSS.

The FRM deals with two major mechanisms:

· local disk, and

· remote Mass Storage System (MSS).

The MSS mechanism is generic and any MSS is able to be used as long as it has a

· Unix-like name space,

· command to transfer data to and from the MSS, and

· command to supply meta-data about files stored in the MSS.

Refer to the section “Interfacing an MSS” for details on how to use an arbitrary Mass Storage System with the FRM.

The FRM is fully compatible with any file system that can be configured to be used with xrootd as it directly uses the oss layer to perform file system operations.

The FRM configuration is described by a configuration file. This configuration file must hold the configuration directives describing the xrootd configuration and well as a few FRM specific directives.

2 frm_admin Command

frm_admin [options] [command [cmdopts] cmdparms]

options: [-c cfn ] [-d] [-h] [-n name] [-v]

q[uery] | quit | reloc | rm

Function

Execute local file residency maintenance commands.

Options

-c cfn The configuration file to be used. The default configuration file, /opt/xrootd/etc/xrootd.cf, is read is it exists.

-d Turns on debugging.

-h Prints usage information.

-n name

Is the name of the server instance for which the command is being performed. See the notes for more information.

-v Prints additional messages where relevant.

command

Is one of the following:

help - provides additional usage information.

audit - verifies that cached file system consistency.

exit - terminates the program (used in interactive mode).

find - finds existing or missing residency management meta-files.

makelf - creates missing lock meta-files or modifies existing ones.

pin - creates or modifies pin meta-files.

query - displays various information.

reloc - relocates files from one type of space to another.

rm - removes one or more files or directories.

cmdopts

Are command specific options. Refer to each command description for more information.

cmdparms

Are command specific parameters. Refer to each command description for more information.

Notes

1) When frm_admin is invoked with no parameters, it enters interactive prompt mode. In this mode you can execute a series of commands by entering new line separated functions and parameters in response to the prompt. Command-line editing tools are enabled on platforms that support them.

2) To leave interactive mode, use the exit or quit commands. Alternatively, a null line in response to the prompt causes the program to exit.

2.1 audit Subcommand

names ldir

audit [options] space sname[:sdir]

usage sname[:sdir]

options: [-fix] [-force] [-migratable] [-purgeable]

[-recursive]

Function

Inventory the logical name space, physical data space, or usage information for validity.

Parameters

names

Audits the logical name space, ldir.

ldir Is the logical name of the directory whose contents are to be audited.

space

Audits the physical data space named sname.

usage

Audits the space usage associated with physical data space named sname.

sname Is the name of the space to be audited for contents or usage. This is the name that is specified using the oss.cache directive. By default, all directories associated with the named space are processed.

sdir Is a specific directory associated with sname, as specified using the oss.cache directive. Only that directory is processed.

Options

-fix Attempts to fix problems encountered during the audit.

-force When used in conjunction with –fix, automatically answers yes to every question posed. This, in effect, corrects all mistakes without intervention.

-migratable

With –fix, creates any missing lock meta-files in ldir indicating that the file associated with the missing lock meta-file must be migrated prior to being purged. This is the default when ldir is exported with the mig option. See the notes for more information on lock meta-files.

-purgeable

With –fix, creates any missing lock meta-files in ldir indicating that the file associated with the missing lock meta-file may be purged. See the notes for more information on lock meta-files.

-recursive

Recursively audits the names in all directories starting with ldir.

Notes on audit names

1) Lock meta-files are special files placed in the logical name space that are used to track file modifications. These files are automatically created when the space is exported with the mig or stage options.

2) When a names audit is requested the following occurs

a. The directory is scanned for any orphaned meta-files. These files have a suffix of “.fail”, “.lock”, “.map”, and “.pin”. An orphaned meta-file is a file that exist without the presence of its associated base (i.e., data) file. The default fix action is to remove orphaned files.

b. If the name space has been exported with the mig or stage options, the audit verifies that each base file has an associated lock meta-file. The default fix action is to create missing lock files that either requires the base file to be migrated or allows it to be purged; depending on the subcommand options or defaults.

c. Should a base file be a symbolic link to a named space, the link’s consistency is verified. If the link point to a non-existent data file); the default fix action is to remove the dangling link.

d. Additionally for XA spaces, the data file must be associated with a pfn meta-file. If the pfn meta-file is missing or incorrect, the default fix action is to create a new one.

Notes on audit space

1) Generally, space auditing is a deterministic process. Occasionally, irresolvable conflicts can arise. You should not use the auto option to fix spaces without first running the audit names subcommand.

2) When a space is audited the following occurs

a. The directory containing the space is scanned to verify that each data file in the space is pointed to by a symbolic link in the associated name space. When a missing link is found and the logical name of the file can be determined, the default fix action is to create the symbolic link. If the logical name cannot be determined, the default fix action is to delete the data file and its associated meta-files.

b. Additionally for XA spaces, the data file must be associated with a pfn meta-file. If the pfn meta-file is missing; the default fix action is to delete the data file. If the pfn meta-file is incorrect, the default fix action is to create a new symbolic link in the logical name space pointing to the data file.

Notes on audit usage

1) Usage auditing calculates the amount of space used by files associated with a particular space. It is only meaningful for XA spaces.

2) The –fix option is only meaningful when usage logging is enabled (see the oss.usage directive).

3) Space usage can be fixed while the xrootd server is running. Applying multiple fixes is allowed.

2.3 makelf Subcommand

makelf [options] lspec [lspec [. . .]]

options: [-migratable] [-owner ospec] [-purgeable]

[-recursive]

ospec: usr[:grp] | [usr]:grp

lspec: lfn | ldir[/*]

Function

Create or alter a lock meta-file.

Parameters

lfn Is the logical name of a file for which a lock meta-file is to be created or altered. Multiple lfn’s may be specified.

ldir Is the logical name of a directory. When ldir is suffixed by “/*’ then lock meta-file are created or altered for all data files in ldir. Specifying –recursive applies makelf to all files in ldir and all of its descendants. Multiple ldir’s may be specified.

Options

-migratable

Creates or alters the lock meta-file to indicate that lfn must be migrated prior to being purged. This is the default when lfn is in a directory exported with the mig option and the lock file does not exist.

-owner

Sets the lock meta-file ownership. The default ownership corresponds to the current effective user and group id. To change the user id, specify usr as a login username or uid. To change the group id, specify grp as a group name or gid. Setting ownership may require root privileges.

-purgeable

Creates or alters the lock meta-file to indicate that lfn may be purged.

-recursive

Recursively processes all subdirectories of ldir.

Notes

1) The makelf command is useful only for name spaces that have been exported with the mig or stage options.

2) You must quote or escape ldir when specifying ldir with an asterisk on the frm_admin command line.

2.4 pin Subcommand

pin [options] lspec [lspec [. . .]]

options: [-keep tspec] [-owner ospec] [-recursive]

tspec: [+]num[d|h|m|s] | mm/dd/[yy]yy | forever

ospec: usr[:grp] | [usr]:grp

lspec: lfn | ldir[/*]

Function

Create or alter a pin meta-file to control file purging.

Parameters

lfn Is the logical name of a file for which a pin meta-file is to be created or altered. Multiple lfn’s may be specified.

ldir Is the logical name of a directory. When ldir is suffixed by “/*’ then pin meta-file are created or altered for all data files in ldir. Specifying –recursive applies pin to all files in ldir and all of its descendants. Multiple ldir’s may be specified.

Options

-keep Sets amount of time a file must be kept in local disk. To keep a file on disk unless it has not been accessed for a period of time, specify the period as +num[d | h | m | s] where num is the quantity and d is for days, h for hours, m for minutes, and s for seconds (the default). To keep a file on disk for a certain amount of time past midnight, specify the time as num[d | h | m | s] without the leading plus sign. To keep a file on disk until a particular date, specify the time in date format (i.e., keep until this date) as mm/dd/[yy]yy where mm is the month, dd is the day, and yy is the year with two digit years being automatically prefixed by 20. To keep the file on disk indefinitely, specify the time as forever.

-owner

Sets the pin meta-file ownership. The default ownership corresponds to the current effective user and group id. To change the user id, specify usr as a login username or uid. To change the group id, specify grp as a group name or gid. Setting ownership may require root privileges.

-recursive

Recursively processes all subdirectories of lspec should it be a directory.

Notes

1) The pin command is useful only for name spaces that have been exported with the mig or stage options.

2) Default pinning attributes are based on the configuration supplied to the purge daemon. The pin meta-file over-rides the defaults on a per-file basis.

3) When keep time is zero, the pin meta-file is removed.

4) You must quote or escape ldir when specifying ldir with an asterisk on the frm_admin command line.

2.5 query Subcommand

pfn lspec [lspec [. . .]]

query rfn lspec [lspec [. . .]]

space [[-recursive] lspec [lspec [...]]

usage [sname]

lspec: lfn | ldir[/*]

Function

Display various information.

Parameters

pfn Displays the local disk physical file name corresponding to the specified logical name. The name need not exist. More than one lspec may be specified.

rfn Displays the remote disk (i.e., MSS) physical file name corresponding to the specified logical name. The name need not exist. More than one lspec may be specified.

space Without arguments, displays the configured spaces as defined by the oss.cache directive. When lspec is specified, displays the space in which one or more files reside.

lfn Is the logical name to be queried. Multiple lfn’s may be specified.

ldir Is the logical name of a directory to be queried. Multiple ldir’s may be specified.

ldir/* Is the logical name of a directory. For query space, the query is applied to all data files in the directory. This is the default for query space when lspec is a directory name. Multiple ldir/*’s may be specified.

usage Displays usage information based on the usage log file for the space named sname. If sname is not specified, usage for all named spaces is displayed.

Options

-recursive

Recursively processes all subdirectories of lspec should it be a directory.

Notes

1) The query space subcommand is only useful when a cached file system has been defined with the oss.cache directive.

2) The query usage subcommand is only useful with usage logging has been enabled with the oss.usage directive.

3) Usage for non-XA spaces contains limited information and is not necessarily accurate relative to other non-XA spaces.

4) Usage information is displayed in bytes with the following tags:

a. Space the name of the space.

b. Used bytes currently recognized by xrootd as being used.

c. Staged bytes staged by the pre-stage daemon.

d. Purged bytes purged by the purge daemon.

e. Adjust the correction factor determined by audit usage –fix.

f. Effective (Used + Staged – Purged + Adjust)

5) The staged, purged, and adjust values are periodically reset to zero when the xrootd daemon re-computes the quantity it considers in-use based on the previous values.

6) You must quote or escape ldir when specifying ldir with an asterisk on the frm_admin command line.

4 The frm_pstga Command for the Pre-Staging Agent

frm_pstga [ options ]

options: [-c cfn] [-d] [-k {num | sz{k|m|g}}] [-l logfn]

[-n name]

Function

Add, delete, and list entries in the frm_pstgd request queue.

Options

-c configfn

The configuration file name. The default name for the configuration file comes from the environmental variable XRDCONFIGFN, if set[1], otherwise it defaults to “/opt/xrootd/etc/xrootd.cf”.

-d Turns on debugging mode. Additional information is printed to describe various actions.

-k num | sz{k|m|g}

Keep no more than num old log files. If sz is specified, the number of log files kept (excluding the current log file) is determined by how much space they use. Hence, kept log files will not exceed sz bytes. The sz must be suffixed by k, m, or g to indicate kilobytes, megabyte, or gigbytes, respectively.

-l logfn

Routes error messages and any trace output to logfn. By default, messages are directed to standard error unless the environmental viable XRDLOGDIR is set[2], in which case the logfn defaults to “$XRDLOGDIR/frm/pstgalog”. Refer to the notes on how logfn is modified by the –n option.

-n name

Assigns name to the frm_pstga instance. The default name comes from the environmental variable XRDNAME, if set[3], otherwise frm_pstga is unnamed. See the notes on how to use this option.

Notes

5) The frm_pstga also accepts additional directives from the configuration file, configfn. These directives are described in the following section.

6) The –n option allows you to run multiple instances of frm_pstga with a common configuration file. This is possible because the specified name is used to modify various file system paths frm_pstga uses for output files. By automatically differentiating such paths by instance name prevents two frm_pstga processes from interfering with each other.

7) The –n option modifies the logfn specified on the command line. Assuming logfn is composed of “/path/fn” and –n name is specified, the log file name becomes “/path/name/fn”.

8) The frm_pstga command reads requests from standard input and provides responses via standard out. This is compatible with the requirements of the xrootd’s oss.stagecmd and oss.stagemsg directives and cmsd’s cms.prep and cms.prepmsg directives. Refer to each respective manual for more information.

9) To enable FIFO notification while using xrootd you must specify

a. frm_pstga as the staging command via the oss.stagecmd directive and

b. Use the async option, and

c. pipe requests into frm_pstga (i.e., specify “|”).

10) The frm_pstga command can also be used with any other program to manage the pre-stage queue. Refer to the description of frm_pstga commands for more information.

4.1 The frm_pstga Configuration directives

In addition to the directives prefixed by “frm.pstg”, frm_pstga also uses certain xrd-specific directives, if specified, for configuration. They are shown in gray below.

all.adminpath apath

frm.pstg.queuepath qpath

Directives

all.adminpath apath

Specifies the fully qualified administrative path where various directories and special files may be created to control execution. The default name for the apath comes from the environmental variable XRDADMINPATH, if set[4], otherwise it defaults to “/tmp”.

frm.pstg.queuepath qpath

Specifies the directory where the pre-stage queue files and other special files reside. The directory path and the files are automatically created if they are missing. The default path is “/apath[/name]/frm” where apath comes from the specified or defaulted adminpath directive and name from the –n command line option.

Notes

1) The frm_pstga communicates with frm_pstgd via a named UDP socket. This socket is located “/apath[/name]/frm/pstg.udp”.

4.2 The frm_pstga Commands

The following commands can be directed to frm_pstga via standard-in:

Add to queue: +[traceid] requestid notify prty mode lfn[?cgi]

Remove from queue: - requestid

List the queue: ? [varnames]

tid | tod | [varnames]

Add to queue (+) parameters

traceid

a 1- to- 256 character identifier describing the entity that caused the entry to be added. See the notes for an explanation of the standard traceid format. A generic default is used if traceid is not specified.

requestid

a 1- to- 64 character identifier to be used to group this request into a unique set of requests. This means that requestid should be globally unique.

notify a 1- to- 512 character string describing who should get notified upon completion of the request. Multiple strings, totaling 512 characters, can be specified by each but the last one with the carriage return (‘\r’) character. See the section on frm_pstgd messages for the actual message that is sent. The notify format[5] is:

- no notification is to be sent.

file:////path send a special message via FIFO named path

udp://rhost:port/msg send msg via udp to rhost:port

udp://path/msg send msg via udp via local Unix path path

prty the character ‘0’, ‘1’, or ‘2’, designating the request’s priority (0 being the lowest). If an invalid priority is specified, ‘0’ is used. See frm_pstgd on how request priorities are used.

mode The processing mode as a sequence of characters:

r[x] - prestage a file in read-only mode. See below for the x values.

w[x] - prestage a file in read-write mode. See below for the x values.

* - end the requestid set and optionally send notification when the

requested set actually completes[6].

The r and w functions can be suffixed with additional letters to indicate how errors and successes are to be handled. As shown above, x can be a combination of the following:

f - send notification if the file cannot be pre-staged. This option is

unaffected by the q letter.

s - send notification upon successfully staging in the file[7].

q - do not send any notifications associated with this entry. By default,

only failure notices are sent.

lfn the logical file name of the file to be staged.

cgi opaque information to be forwarded to the staging transfer agent, if configured to accept this information (see frm_pstgd’s frm.pstg.xfrcmd configuration directive). The total amount of information, including the lfn and ‘?’ separator, cannot exceed 2,175 characters.

Remove from queue (-) parameters

requestid

a 1- to- 64 character identifier to be used to identify the requests that are to be removed. All pending requests associated with this identifier are removed. Requests already in progress remain are allowed to complete even though they are no longer in the queue.

List the queue (?) parameters

varnames

is an optional list of variable names, as described below. Information is listed in the order in which the variables are listed; each separated by a space. The default variable is lfn. Invalid varnames are silently ignored. Valid variables are:

Var	Information	Var	Information	Var	Information
lfn	Logical File Name	note	Notification string	rid	requestid
lfncgi	lfn?[cgi string]	prty	Priority	tid	traceid
mode	Processing mode	qwt	Seconds in queue	tod	Time of day added

5 frm_pstgd Command for the Pre-Staging Daemon

frm_pstgd [ options ]

options: [-c cfn] [-d] [-k {num | sz{k|m|g}}] [-l logfn]

[-n name] [-s]

Function

Manage the copying of files from a Mass Storage System to local disk.

Options

-c configfn

The configuration file name. The default name for the configuration file is “/opt/xrootd/etc/xrootd.cf”.

-d Turns on debugging mode. Additional information is printed to describe various actions.

-k num | sz{k|m|g}

-l logfn

Routes error messages and any trace output to logfn. By default, messages are directed to standard error.

-n name

Assigns name to the frm_pstgd instance. By default, the frm_pstgd instance is unnamed. See the notes on how to use this option.

-s Logs staging statistics to the log file or standard error.

Notes

1) The frm_pstgd also accepts additional directives from the configuration file, configfn. These directives are described in the following section.

5.1 The frm_pstgd Configuration directives

In addition to the directives prefixed by “frm.pstg”, frm_pstgd also uses certain ofs, oss, and xrd-specific directives, if already specified, for configuration. They are shown grayed-out below.

all.adminpath apath

frm.pstg.stopfile sfn

frm.pstg.queuepath qpath

frm.pstg.waittime wsec

frm.pstg.xfrcmd cmd [ args ]

frm.pstg.xfrmax xmax

ofs.osslib libpath [ parms ]

oss.localroot lpath

oss.namelib npath [ parms ]

oss.remoteroot rpath

xrootd.monitor dest stage host:port

―――――――――――――――――――――――――――――――――――――――――――――――――――――――――――――――――――

args: [text] [var] [args]

var: $CGI | $LFN | $PFN | $RFN | $NOTIFY | $OFLAG |

$PRTY | $RID | $USER | $eVar

Directives

all.adminpath apath

Specifies the fully qualified administrative path where various directories and special files may be created to control execution. The default is “/tmp”.

frm.pstg.stopfile sfn

Specifies the fully qualified “stop” file name. When sfn is present, new pre-stage requests are deferred until the file is no longer present. This effectively provides the ability to suspend pre-staging without having to kill the daemon. The default is “/apath/frm[/name]/STOPPSTG””, where apath comes from the adminpath directive and name from the –n command line option.

frm.pstg.queuepath qpath

Specifies the directory where the pre-stage queue files and other special files reside. The directory path and the files are automatically created if they are missing. The default path is “/apath/frm[/name]” where apath comes from the adminpath directive and name from the –n command line option.

frm.pstg.waittime wsec

wsec specifies the number of seconds to wait before checking the pre-stage queue for outstanding requests in an otherwise idle system. The default is 300 seconds. See the usage notes before changing the default.

frm.pstg.xfrcmd cmd [ args ]

Specifies the absolute path of the program that frm_pstgd is to use to transfer a file from the remote MSS to local disk. The default command is:

“/opt/xrootd/utils/frm_xfr –p $OFLAG $RFN $PFN”

The args are arbitrary command-specific tokens and may include special variables that will be substituted each time the command is invoked (as used in the default command). The following variables may be specified:

$CGI all of the opaque information specified after the question mark in the file path

$LFN logical file name

$PFN physical file name as modified by localroot or the namelib

$RFN remote MSS file name as modified by remoteroot or the namelib

$NOTIFY notification string. See the cms.prepmsg directive for details.

$OFLAG a character letter describing the file open processing flags:

w – O_WRONLY | O_RDWR r – O_RDONLY

$OPTS notification and other options (see cms.prepmsg directive)

$PRTY request priority

$RID request identifier

$TID original requesting client’s trace identification

$eVar any variable that has been passed along with the file name as opaque information (i.e., contained in $CGI).

frm.pstg.xfrmax xmax

xmax specifies the maximum number of files that may be transferred at the same time. The default is 1.

ofs.osslib libpath [ parms ]

Specifies the absolute path to the shared library that contains the implementation of the storage system interface that frm_pstgd is to use for file system specific storage operations (e.g., create, rename, etc). The default is to use a built-in mechanism that is identical to what xrootd uses by default.

oss.localroot lpath

The path that is to be used as a prefix to any logical file name before using it as a local physical file name (i.e. to map lfn to local pfn). By default, no prefix is used. Also see the oss.namelib directive.

oss.namelib npath [ parms ]

Specifies the absolute path to the shared library that contains the implementation of the name-to-name translation interface that frm_pstgd is to use to convert logical file names to local and remote physical file names. The default is to use a built-in mechanism that relies on the oss.localroot and oss.remote root directives.

oss.rmoteroot rpath

The path that is to be used as a prefix to any logical file name before using it as a remote physical file name (i.e. to map lfn to Mass Storage System pfn). By default, no prefix is used. Also see the oss.namelib directive.

5.2 The frm_pstgd Notification Messages

When frm_pstgd is asked to send a notification it uses one or two message formats as shown in each diagram below.

Successful: stage OK lfn

Unsuccessful stage {ENOENT | BAD} lfn [ emsg ]

Notification: file:////path

Where

lfn logical filename of the file that successfully staged in or whose staging failed.

ENOENT

staging failed because the file could not be found.

BAD staging failed for other reasons (see error_msg).

emsg an optional error message describing the nature of the failure.

Successful: avail requestid msg lfn

Unsuccessful: unavail requestid msg lfn

Notification: udp://rhost:port/msg or udp://path/msg

Where

requestid

the requestid associate with the staging request.

msg the message, if any, present in the notification string.

lfn the logical filename of the file that successfully staged in or whose staging failed.

6 The frm_xfr.hpss Transfer Command

frm_xfr.hpss [ options ] [user@host:]sfn [user@host:]tfn

options: [-d] [-f] [-k keytab] [-m] [-n] [-p {r|w|mode}]

[-P port] [-t tries] [-x pftpcmd] [-z]

Function

Copy a file to local disk; typically, from a Mass Storage System.

Parameters

[user@host:]sfn

name of the source file to read. If it is prefixed by “user@host:”, then the file must exist in a pftp accessible Mass Storage system located at host and reachable via login as user. Otherwise, the file whose name is sfn must be present on local disk.

[user@host:]tfn

name of the source file to write. If it is prefixed by “user@host:”, then the file is to be written into a pftp accessible remote Mass Storage system located at host and reachable via login as user. Otherwise, the file will be written to local disk using the name sfn.

targetfn

The name of the file as it should exist on local disk.

Options

-d Turns on debugging.

-f over-writes any local file, even if it has zero length. The –f option is the default for zero-length local files. It has no effect on remote files.

-k keytab

specifies the pftp password’s location. The keytab must contain a single password that will be used when logging in as user@host. The default location is “/var/adm/xrootd/pftp/keyfile”.

-m when tfn is local, verifies that tfn does not exist. Should tfn exist, the command fails with an error message.

-n does not redirect standard error to standard out when invoking the xfrcmd. This is meant purely for debugging purposes.

-p {r | w | mode}

sets the permission bits for the target file. ‘r’ implies a mode of 0440, ‘w’ a mode of 0640, and mode is 1- to 4-character octal mode. The default mode is determined by the xfrcmd that is used to transfer the file.

-P port

is the port number to contact when logging in using user@host. The default port number is 2021.

-t tries

The number of times to try a failing copy when the error indicates a transient condition. The default is 2.

-x xfrcmd

is the pftp command to be used to copy sfn to tfn. The default is to use “/opt/xrootd/utils/pftp_client”.

-z does not establish a new process group prior to executing the xfrcmd.

6.1 The frm_xfr.hpss pftp Command Streams

user user pswd

binary

setpblocksize 2097152

site setcos cos

mkdir dirpath

●

pput localfn mssfn

quit

Transfer Command Stream When Sending a File to the MSS

Where:

user The user contained in “user@host:tfn” for the target file.

keyfn The password in the file identified by –k command line option.

cos The value specified with the –s command line option.

dirpath The path segment that needs to exist in order to store the file. There are as many mkdir commands as there are path segments to ensure that the complete target path exists in the MSS.

localfn The name of the file to be placed in the MSS as it is known on local disk.

mssfn The name of the file as it must be stored in the MSS.

user xfruser pswd

binary

setpblocksize 2097152

pget mssfn localfn

quit

Transfer Command Stream When Retrieving a File from the MSS

Where:

user The user contained in “user@host:tfn” for the target file.

keyfn The password in the file identified by –k command line option.

mssfn The name of the file as it exists in the MSS.

localfn The name of the file as it must be created on local disk.

[1] The cmsd and xrootd set this environmental variable if passed the –c command line option.

[2] The cmsd and xrootd set this environmental variable if passed the –l command line option.

[3] The cmsd and xrootd set this environmental variable if passed the –n command line option.

[4] The programs cmsd and xrootd set this environmental variable.

[5] The old mps_prep and mps_PreStage “mailto://” and “tcp://” notify options are not supported.

[7] The letter ‘n’ is a synonym for ‘s’.

1 Introduction

1.1 FRM Components

2 frm_admin Command

2.1 audit Subcommand

2.2 find Subcommand

2.3 makelf Subcommand

2.4 pin Subcommand

2.5 query Subcommand

2.6 reloc Subcommand

2.7 rm Subcommand

3 The frm_migrd Command for Migration

4 The frm_pstga Command for the Pre-Staging Agent

4.1 The frm_pstga Configuration directives

4.2 The frm_pstga Commands

5 frm_pstgd Command for the Pre-Staging Daemon

5.1 The frm_pstgd Configuration directives

5.2 The frm_pstgd Notification Messages

6 The frm_xfr.hpss Transfer Command

6.1 The frm_xfr.hpss pftp Command Streams

7 The frm_purgd Command for Purging

8 Document Change History