iRods for HUBs

We are working on using iRODS, as a way to increase storage for users of the HUB. Possible uses include simple staging of data for simulation, and housing searchable collections of documents. Some theory about our approach can be found in the paper here.

We were using two servers to test iRods at CCR. The first was both a data and metadata server. The second was just a data server. Now, we upgraded to iRODS 3.0 and are only using one server.

Setup for New User

Ask the administrator to create an irods account for you with a password you’d like.

File: irodsEnv (435 B , uploaded by Kyle Marcus 1 decade 2 years ago)

Modify the user name inside the file linked above, and change the file name to .irodsEnv

Create a folder name .irods in the user’s home directory and place .irodsEnv in the folder.

Create a folder named irods in your home directory. This folder will be used as the mount location for the the iRODS filesystem.

Type the command use irods in your workspace command line to be able to run icommands.

iRods.png

To make your password persist, type iinit and enter your irods password.

Create a bash profile file name .bash_profile with the following content:

if [ -n $IRODS_MOUNT -a -e $IRODS_MOUNT ]; then
$IRODS_MOUNT
fi

Then, run the command source .bash_profile

Finally, if you type ls irods from your home directory, you should see all the users’ home folders in the irods directory, including your own.

.bashrc files for users

Once icommands and irodsFs are installed into /apps the script will read similar to this:

PATH=/apps/iRODS/clients/icommands/bin:/apps/iRODS/clients/fuse/bin:$PATH
irodsFs $HOME/irods 1>/dev/null 2>&1

submit and irods

Make sure preprocess_irods.sh and postprocess_irods.sh are set to executable

chmod +x preprocess_irods.sh
chmod +x postprocess_irods.sh

Also when using the submit command, be sure to include the preprocess_irods.sh and postprocess_irods.sh files

submit ... -i preprocess_irods.sh -i postprocess_irods.sh ...

If you are placing the output of a job into iRods, make sure to use the -q option with submit so that there will be no limit on the file size that you can create.

modification to /san/user/vhub/u2/bin/receiveinput.sh

IRODS_INPUT_NAME="preprocess_irods.sh"
if  [ -x ./$IRODS_INPUT_NAME ]; then
./$IRODS_INPUT_NAME
fi

modification to /san/user/vhub/u2/bin/transmitresults.sh

IRODS_OUTPUT_NAME="postprocess_irods.sh"
if [ -x ./$IRODS_OUTPUT_NAME ]; then
./$IRODS_OUTPUT_NAME
fi

Common Problems

Client cannot connect to server

Server may go down unexpectedly. Admin should log onto irods1.ccr.buffalo.edu, change to the iRods directory and type

irodsctl restart

It’s better to use restart than istart because sometimes the SQL server has gone down but the iRODS daemon is still running. istart will not resart the database.

Check the server log to see whether the restart succeded. The log is in iRODS/server/log and it is named by the date, e.g. rodsLog.2010.7.11

vhub cannot access the users file

The user must do the command

ichmod inherit username

ichmod -r own vhub /vhub/home/username

Idea for submitted apps to get irods files

Use file browser to get path to file.

  • Parse path to get file name.
  • when submit arrives, it makes a symbolic link to the file name and name the link with the file name.
  • remove the symbolic link after the app runs.
ln -s irods/username/pathToFile/filename filename
  • submit execs app, file “resides” in cwd.
rm filename

To use a remote command-line client:

Get a copy of the file .irodsEnv from us (Note: the irodsHost needs to be an IP address rather than a domain name) and have us set up your login name and password on the iRods server.

From your home directory on the hub:

mkdir ~/.irods
mv .irodsEnv ~/.irods

Download iRods

Follow these instructions for installing additional clients.

At this point you will need to set your path:

PATH=home/vhub/username/iRODS/clients/icommands/bin:$PATH

It’s easiest if you put this command in your .bashrc or .profile file so the commands are available every time you open a terminal.

Some basic concepts

  • collection: a directory
  • resource: represents a physical data server location that can hold many collections
  • zone: multiple resources can be members of the same zone

Some useful commands

  • iinit: enter your password and it will get cahed so you don’t have to give your password for every command.
  • ils: list files
  • imkdir: create a directory (also called a collection)

imkdir username/folder

  • iput: put a file into iRods

iput foo.tar username/folder

  • ireg: copy a directory revursively into iRods

NOTE: If you add files to the actual physical directory, this is not reflected in iRODS.

ireg -C /data/iRODS/testIreg/ /tempZone/home/rods/newCol
ils newCol
/tempZone/home/rods/newCol:
   foo1
   foo2
cd testIreg/
touch bar
ils newCol
/tempZone/home/rods/newCol:
   foo1
   foo2

iRods FUSE notes

Install: Need to install libfuse-dev to get the header files in the same path as the library

edit iRODS/config/config.mk

uncomment IRODS_FS = 1

set fuseHomeDir=path/to/fuse where the path is just above include/fuse.h and lib/libfuse.a

on your vhub account this is simply fuseHomeDir=/usr


in iRODS/client/fuse do make

Add another path so that you can mount with the command irodsFs

NOTE: You cannot unmount on your vhub account, once you’ve mounted irods. Only a vhub dministrator can run fusermount.

PATH=/home/alisanee/iRods/iRODS/clients/icommands/bin:/home/alisanee/iRods/iRODS/clients/fuse/bin:$PATH

Public share directory

While using iRODS you may want to share files with different users. One way to do this is to put the files you want to share in /vhub/home/public. Any user that is on the iRODS system can read and write to this public directory.

icp foo /vhub/home/public

ADMIN NOTE: When users are created, they are automatically added to the share group. The permissions on the /vhub/home/public directory is set up so anyone in the share group is allowed r/w access.

If you want to allow users that do not have an account on the iRODS system to access the public share directory, you need to create the anonymous user. Create the user just like all the other users, but you do not have to set the password.

iadmin mkuser anonymous rodsuser

After creating the user, allow that user to read the /vhub/home/public directory.

ichmod read anonymous /vhub/home/public

ichmod inherit anonymous

After these steps, anyone that does not have an account on the irods system is able to read the files located in the /vhub/home/public directory. They are not able to write or go up the directory structure.

In order for users to login to iRODS with this anonymous user, all you have to do is use the username anonymous with any password.

The username anonymous is a special account name that is recognised in the iRODS system.

Admin’s can use the following script to automate the process of creating the necessary files needed to use the anonymous user in vhub and mount the /vhub/home/public directory on iRODS to ~/irods directory on the users filesystem on vhub.

File : irods_anonymous.sh (4 KB, uploaded by Kyle Marcus 1 decade 2 years ago)

Administrators’ To Do List For Installing on HUB

put the icommands in the path for every user

install .irods directory for every user and provide an irodsEnv file with their Hub ID as the user name

do iinit once with their user name and password. This will create a file, .irods/.irodsA which holds the user’s encrypted password, so that they won’t have to enter their password to do an icoomand.

in irods, create a home directory for the user and set the group to the user name.

irodsFs mount in an empty folder named irods

irodsFs irods

For New User

As the ‘rods’ administrator, for a new hub user:

iadmin mkuser username#vhub rodsuser create new user (use rodsadmin if they are an admin)

iadmin moduser username password somepassword set user’s password

imkdir username make home directory for new user

ichmod -V own username username let user rwx her home directory

ichmod read vhub /home/username let vhub rwx user’s home directory

ichmod write vhub /home/username

ichmod inherit username let vhub rwx the subdirectories and files the user creates

NOTE: if you are creating a rodsadmin account, then you must use the -M option with ichmod

users can then do

iput

iput foo username

Note that the irods1.ccr.buffalo.edu name is not available to a DNS server, so until that changes, the irodsEnv has to be edited to have the IP address instead.

Edit .bashrc in iRods server(done):

PATH=/data/iRODS/iRODS/clients/icommands/bin:$PATH

Install iRods on vhub@u2-grid (done)

edit .bashrc

PATH=/san/user/vhub/u2/iRODS/clients/icommands/bin:$PATH

iadmin mkuser username rodsuser

iadmin mkuser vhub rodsadmin

Transferring large files

When you send or receive a file to/from iRODS that’s over 32 MB, it starts opening multiple ports so it can multithread the data transfers. You need to make sure the correct ports are open or the data transfer will not work.

Open ports 20,000 to 20,199 in the firewall

For irods@irods1.ccr.buffalo.edu edit ~/iRODS/config/irods.config to say

$SVR_PORT_RANGE_START = '20000';
$SVR_PORT_RANGE_END = '20199';

and also

~/iRODS/scripts/perl/irodsctl.pl

to say

$svrPortRangeStart=20000;
$svrPortRangeEnd=20199;

Then restart the server

irodsctl restart

Note: when doing the multithreaded data transfers we had a problem with the server only listening on the internal IP address.

> iput -V -f test_dir_timing_40mb_80files.tar kmarcus/test_time
NOTICE: irodsHost=irods1.ccr.buffalo.edu
NOTICE: irodsPort=1247
NOTICE: irodsDefResource=demoResc
NOTICE: irodsHome=/vhub/home
NOTICE: irodsCwd=/vhub/home
NOTICE: irodsUserName=kmarcus
NOTICE: irodsZone=vhub
>From server: NumThreads=2, addr:10.6.1.20, port:20169,
cookie=810769104
ERROR: connectToRhostPortal: connectTo Rhost 10.6.1.20 port 20169
error, status = -347000

Fix: edit ~/iRODS/server/config/irodsHost to say:

localhost irods1.ccr.buffalo.edu 10.6.1.20

This tells iRODS to prefer the interface irods1.ccr.buffalo.edu when binding.

Information for this fix was provided by Chris Smith on iRODS Chat

Uninstalling iRODS server

remove the collections (i.e. directories and files), e.g.

irm -r aneeman

turn irods off

irodsctl stop

remove directories for irods and its postgres installation.

Changing the server

Rename resource

iadmin modresc oldName name newName

edit config/irods.config

$RESOURCE_NAME = 'newName';

restart the server

./irodsctl restart

Change hostname

stop the server

./irodsctl stop

Edit the following files, changing the host name to the new address:

  • .odbc.ini in the home directory
  • .irods/.irodsEnv in home directory
  • pgsql/etc/odbc.ini in the Postgres directory
  • server/config/server.config in iRODS distribution directory

Edit server/config/irodsHost

localhost new-host-name 10.122.1.8

Start iRods again

./irodsctl start

Finally, modify the address of the local resource

iadmin modresc demoResc host new-host-name

Move a collection or file to a specific server

iphymv -r resource2Collection -R demoResc2

-r recursive -R resource name

The “mount” always looks the same, regardless of the physical location. You can move a file to any data server.

iRods Timing

irods_timing.png

Created on , Last modified on