Design:
http://confluence.slac.stanford.edu/display/IEPM/FTMA+Design
Source:
Pre-Req:
You might need to install the following perl modules if not already installed. Easiest way to do this is to use cpan e.g. "cpan module_name"
Perl:
- Module::Load
- HTTP::Daemon
- XML::SAX
- Config::General
- aliased
- Readonly
- Term::ReadKey
- DBI::DBD
- DBD::mysql
- DBD::SQLite
- Class::Accessor
- Class::Fields
- Params::Validate
- Statistics::Descriptive
- Data::UUID
- IO::Interface
- DateTime
- Error
- Date::Manip (this may require perl upgrade to 5.10 or above, directions)
- Log::Dispatch::FileRotate
- Log::Log4perl
Additionally more components may be required, which may come up when one tries to run the startup script of the service (section "How to run the service"):
- XML::LibXML
Installation:
Theoretically, the service can work with any kind of File Transfer protocol. For test purposes and limiting the large amount of possibilities we use GridFTP as a primary File Transfer Tool.
GridFTP Installation Instruction
- There are a number of online resources, that can help an administrator, install GridFTP. Some of them are:
- A bug in GridFTP causes intervals to be calculated incorrectly unless you use a very recent version of GridFTP or the subversion trunk version of NetLogger.
- There is generally no restriction on the version of GridFTP to be used, however, it is advised to use the latest one, with minimum security risks.
NetLogger Installation
http://confluence.slac.stanford.edu/display/IEPM/Installing+NetLogger
PerfSONAR_PS FTMA
Download the Source Code as a tar ball
Code Block tar \-xvf FileTransfer_MA.1b.tar
- mkdir /var/log/perfsonar
- Run the SQL Procedure(mysql) on the netlogger database to produce a secondary database. This secondary database is the one, the FTMA service will be interacting with.
From the svn
- Checkout the latest version of FTMA:
- With username and password:
ORCode Block svn checkout https://svn.internet2.edu/svn/perfSONAR-PS/branches/FileTransfer/
- Anonymously:
Code Block svn checkout https://anonsvn.internet2.edu/svn/perfSONAR-PS/branches/FileTransfer/
Configuration
- Create the log Directory
Code Block mkdir /var/log/perfsonar (Create the log directory as is shown by daemon_logger.conf)
- Create a mysql database for FTMA service to interact with:
Code Block mysql -u $USER -p $PASS -e "create database ft_ma"
- Run the SQL Procedure(mysql) on the netlogger database to produce a secondary database.
This secondary database is the one, the FTMA service will be interacting with. This procedure has the ability to keep checkpoints, so it is supposed to be run, after a certain amount of time has passed. Ideally this can be run every hour or day, to convert NetLogger data, to be available to the user using FTMA service.
Basic Idea behind this procedure, is to convert GridFTP logs from the schema, as is stored by NetLogger to a schema, defined at the start of this page. In order to do so we have two procedures, a master procedure called make_FTDB. This procedure handles the checkpoints, so that the script, starts, from the point, where it stopped on last execution. This procedure also calls the conversion procedure called build_FTDB, with eventID of the tuple to be converted. build_FTDB, picks up all the relevant attributes, from attr table and then inserts the values into EndPoints table, Metadata table, MetaEvent table and Data table. The procedure also creates table if they don't exsist.
Settings for bin/daemon.conf file
Change "db_name", "db_username" and "db_password" in the following file:
Code Block | ||||||||
---|---|---|---|---|---|---|---|---|
| ||||||||
max_worker_lifetime 360
max_worker_processes 30
disable_echo 0
ls_registration_interval 60
ls_instance http://localhost:9995/perfSONAR_PS/services/hLS
root_hints_url http://www.perfsonar.net/gls.root.hints
<port 9000>
<endpoint /perfSONAR_PS/services/FT/MA>
service_type MA
module perfSONAR_PS::Services::MA::FT
<ftma>
service_description FT MA
service_accesspoint http://localhost:9000/perfSONAR_PS/services/FT/MA
enable_registration 0
service_name perfSONAR_PS FT MA
ls_registration_interval 60
service_timeout 360
query_size_limit 100
db_host localhost
db_username root
db_name ft_ma
db_password ****
db_type mysql
</ftma>
</endpoint>
</port>
|
How to run the Service:
Before running the service, we must have the database. So far we only created the database and did not copy any tables/contents. To do this we must have a file that performs sql functions of adding and filling tables. The file should be a part of the FTMA package retrieved via svn checkout and should be placed inside "FileTransfer/perfSONAR_PS-FileTransfer/contrib" directory. The file can used in the following manner (you might need to do sudo):
If mysql not started then start it:
Code Block |
---|
/etc/init.d/mysqld start
|
Load the mysql dump into a database:
Code Block |
---|
mysql -D ft_ma < mysql_backup
|
Once done do the following to start the service:
Code Block |
---|
* cd contrib
* sh FT_MA_Startup.sh \--skip-input || \--help
|
- Attempts to install the missing, perl modules.
- Stops any previous running instance of the service.
- Creates a backup of any existing log file.
- Everytime the script executes, it copies the log file from the main log directory to an hourlyfolder in the same directory. The folder is named as: Ftp_MM-DD-YY:HH
- Starts a new Instance of the service by running daemon.pl
Client Application:
The service includes a client tool to do some preliminary testing. This tool is present inside the bin dir.
Code Block perl FT_client.pl
This simple execution will fetch all the metadata from the service, and provide the user with a final output showing all the metadata keys mapped on source and destination ip addresses.
Code Block perl FTP_client.pl --help -d Switch to debug level, one of 0,1 or 2 --debug Same as -d --url Url to the MA Service(FT) default is localhost --data Output Data as well --src source ip (string) --dst destination ip (string) --SrcPath metadata param: Source file path(string) --DestPath metadata param: Destination file path(string) --stripes metadata param: number of stripes --buffer metadata param: bufer size --block metadata param: block size --streams metadata param: number of streams --program metadata param: program used for file transfer --user metadata param: user, who requested the file transfer --initEpochTime initial Time limit in Epoch (integer) --finalEpochTime Final Time limit in Epoch (integer) --initUtcTime initial Time limit in UTC (string) --finalUtcTime final Time limit in UTC (string) --startTuple start point of results to return --tupleLimit number of tuples to return -h Print this help --help Same as -h
Code Block perl FTP_client.pl -d --data --DestPath=/ --stripes=1 --src=192.168.117.128 --dst=131.225.107.12 --initEpochTime=1220000000 --finalEpochTime=1225408002
o This execution will fetch metadata where destination path, stripes, source ip address and destination ip address is defined by the parameters passed.
o The --data argument makes sure that the returned metadata is then used to fetch data with start time and end time within the given limitation.
o The -d argument provides the debug data as well.
Code Block FT_client.pl -d 2 --data --stripes=1 --user=dang --startTuple=400 --tupleLimit=20
Gliffy Diagram | ||||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|
|
...