

Installing an XRootD Storage Element

EL7 version compatibility

There is an incompatibility with EL7 < 7.5 due to an issue with the globus-gsi-proxy-core package

XRootD is a hierarchical storage system that can be used in a variety of ways to access data, typically distributed among actual storage resources. One way to use XRootD is to have it refer to many data resources at a single site, and another way to use it is to refer to many storage systems, most likely distributed among sites. An XRootD system includes a redirector, which accepts requests for data and finds a storage repository — locally or otherwise — that can provide the data to the requestor.

Use this page to learn how to install, configure, and use an XRootD redirector as part of a Storage Element (SE) or as part of a global namespace.

Before Starting

Before starting the installation process, consider the following points:

• User IDs: If it does not exist already, the installation will create the Linux user ID xrootd
• Service certificate: The XRootD service uses a host certificate at /etc/grid-security/host*.pem
• Networking: The XRootD service uses port 1094 by default

As with all OSG software installations, there are some one-time (per host) steps to prepare in advance:

• Ensure the host has a supported operating system
• Obtain root access to the host
• Prepare the required Yum repositories
• Install CA certificates

Installing an XRootD Server

An installation of the XRootD server consists of the server itself and its dependencies. Install these with Yum:

[email protected] # yum install xrootd

Configuring an XRootD Server

An advanced XRootD setup has multiple components; it is important to validate that each additional component that you set up is working before moving on to the next component. We have included validation instructions after each component below.

Creating an XRootD cluster

If your storage is spread out over multiple hosts, you will need to set up an XRootD cluster. The cluster uses one "redirector" node as a frontend for user accesses, and multiple data nodes that have the data that users request. Two daemons will run on each node:

xrootd
The eXtended Root Daemon controls file access and storage.

cmsd
The Cluster Management Services Daemon controls communication between nodes.

Note that for large virtual organizations, a site-level redirector may actually also communicate upwards to a regional or global redirector that handles access to a multi-level hierarchy. This section will only cover handling one level of XRootD hierarchy.

In the instructions below, <RDRNODE> will refer to the redirector host and <DATANODE> will refer to the data node host. These should be replaced with the fully-qualified domain name of the host in question.

Modify /etc/xrootd/xrootd-clustered.cfg

You will need to modify the xrootd-clustered.cfg on the redirector node and each data node. The following example should serve as a base configuration for clustering. Further customizations are detailed below.

all.export /mnt/xrootd stage
set xrdr = <RDRNODE>
all.manager $(xrdr):3121

if $(xrdr)
  # Lines in this block are only executed on the redirector node
  all.role manager
else
  # Lines in this block are executed on all nodes but the redirector node
  all.role server
  cms.space min 2g 5g
fi

You will need to customize the following lines:

Configuration Line	Changes Needed
all.export /mnt/xrootd stage	Change /mnt/xrootd to the directory to allow XRootD access to
set xrdr=<RDRNODE>	Change to the hostname of the redirector
cms.space min 2g 5g	Reserve this amount of free space on the node. For this example, if space falls below 2GB, xrootd will not store further files on this node until space climbs above 5GB. You can use k, m, g, or t to indicate kilobyte, megabytes, gigabytes, or terabytes, respectively.

Further information can be found at https://xrootd.slac.stanford.edu/docs.html

Verifying the clustered config

Start both xrootd and cmsd on all nodes according to the instructions in the managing services section.

Verify that you can copy a file such as /bin/sh to /mnt/xrootd on the server data via the redirector:

[email protected] # xrdcp /bin/sh  root://<RDRNODE>:1094///mnt/xrootd/second_test
[xrootd] Total 0.76 MB  [====================] 100.00 % [inf MB/s]

Check that the /mnt/xrootd/second_test is located on data server <DATANODE>.

(Optional) Adding Simple Server Inventory to your cluster

The Simple Server Inventory (SSI) provide means to have an inventory for each data server. SSI requires:

• A second instance of the xrootd daemon on the redirector
• A "composite name space daemon" (XrdCnsd) on each data server; this daemon handles the inventory

As an example, we will set up a two-node XRootD cluster with SSI.

Host A is a redirector node that is running the following daemons:

1. xrootd redirector
2. cmsd
3. xrootd - second instance that required for SSI

Host B is a data server that is running the following daemons:

1. xrootd data server
2. cmsd
3. XrdCnsd - started automatically by xrootd

We will need to create a directory on the redirector node for Inventory files.

[email protected] # mkdir -p /data/inventory
[email protected] # chown xrootd:xrootd /data/inventory

On the data server (host B) let's use a storage cache that will be at a different location from /mnt/xrootd.

[email protected] # mkdir -p  /local/xrootd
[email protected] # chown xrootd:xrootd /local/xrootd

We will be running two instances of XRootD on <HOST A>. Modify /etc/xrootd/xrootd-clustered.cfg to give the two instances different behavior, as such:

all.export /data/xrootdfs
set xrdr=<HOST A>
all.manager $(xrdr):3121
if $(xrdr) && named cns
      all.export /data/inventory
      xrd.port 1095
else if $(xrdr)
      all.role manager
      xrd.port 1094
else
      all.role server
      oss.localroot /local/xrootd
      ofs.notify closew create mkdir mv rm rmdir trunc | /usr/bin/XrdCnsd -d -D 2 -i 90 -b $(xrdr):1095:/data/inventory
      #add cms.space if you have less the 11GB
      # cms.space options https://xrootd.slac.stanford.edu/doc/dev410/cms_config.htm
      cms.space min 2g 5g
fi

The value of oss.localroot will be prepended to any file access.
E.g. accessing root://<RDRNODE>:1094//data/xrootdfs/test1 will actually go to /local/xrootd/data/xrootdfs/test1.

Starting a second instance of XRootD on EL 6

The procedure for starting a second instance differs between EL 6 and EL 7. This section is the procedure for EL 6.

Now, we have to change /etc/sysconfig/xrootd on the redirector node (<HOST A>) to run multiple instances of XRootD. The second instance of XRootD will be named "cns" and will be used for SSI.

XROOTD_USER=xrootd 
XROOTD_GROUP=xrootd 
XROOTD_DEFAULT_OPTIONS="-k 7 -l /var/log/xrootd/xrootd.log -c /etc/xrootd/xrootd-clustered.cfg"
XROOTD_CNS_OPTIONS="-k 7 -l /var/log/xrootd/xrootd.log -c /etc/xrootd/xrootd-clustered.cfg"
CMSD_DEFAULT_OPTIONS="-k 7 -l /var/log/xrootd/cmsd.log -c /etc/xrootd/xrootd-clustered.cfg"
FRMD_DEFAULT_OPTIONS="-k 7 -l /var/log/xrootd/frmd.log -c /etc/xrootd/xrootd-clustered.cfg"
XROOTD_INSTANCES="default cns"
CMSD_INSTANCES="default" 
FRMD_INSTANCES="default" 

Now, we can start XRootD cluster executing the following commands. On redirector you will see:

[email protected] # service xrootd start 
Starting xrootd (xrootd, default): [ OK ]
Starting xrootd (xrootd, cns): [ OK ]
[email protected] # service cmsd start 
Starting xrootd (cmsd, default): [ OK ]

On redirector node you should see two instances of xrootd running:

[email protected] # ps auxww|grep xrootd
xrootd 29036 0.0 0.0 44008 3172 ? Sl Apr11 0:00 /usr/bin/xrootd -k 7 -l /var/log/xrootd/xrootd.log -c /etc/xrootd/xrootd-clustered.cfg -b -s /var/run/xrootd/xrootd-default.pid -n default
xrootd 29108 0.0 0.0 43868 3016 ? Sl Apr11 0:00 /usr/bin/xrootd -k 7 -l /var/log/xrootd/xrootd.log -c /etc/xrootd/xrootd-clustered.cfg -b -s /var/run/xrootd/xrootd-cns.pid -n cns
xrootd 29196 0.0 0.0 51420 3692 ? Sl Apr11 0:00 /usr/bin/cmsd -k 7 -l /var/log/xrootd/cmsd.log -c /etc/xrootd/xrootd-clustered.cfg -b -s /var/run/xrootd/cmsd-default.pid -n default

Warning

The log file for second named instance of xrootd with be placed in /var/log/xrootd/cns/xrootd.log

On data server node you should that XrdCnsd process has been started:

[email protected] # ps auxww|grep xrootd
xrootd 19156 0.0 0.0 48096 3256 ? Sl 07:37 0:00 /usr/bin/cmsd -l /var/log/xrootd/cmsd.log -c /etc/xrootd/xrootd-clustered.cfg -b -s /var/run/xrootd/cmsd-default.pid -n default
xrootd 19880 0.0 0.0 46124 2916 ? Sl 08:33 0:00 /usr/bin/xrootd -l /var/log/xrootd/xrootd.log -c /etc/xrootd/xrootd-clustered.cfg -b -s /var/run/xrootd/xrootd-default.pid -n default
xrootd 19894 0.0 0.1 71164 6960 ? Sl 08:33 0:00 /usr/bin/XrdCnsd -d -D 2 -i 90 -b fermicloud053.fnal.gov:1095:/data/inventory

Starting a second instance of XRootD on EL 7

The procedure for starting a second instance differs between EL 6 and EL 7. This section is the procedure for EL 7.

1. Create a symlink pointing to /etc/xrootd/xrootd-clustered.cfg at /etc/xrootd/xrootd-cns.cfg:

[email protected] # ln -s /etc/xrootd/xrootd-clustered.cfg /etc/xrootd/xrootd-cns.cfg

1. Start an instance of the xrootd service named cns using the syntax in the managing services section:

[email protected] # systemctl start [email protected]

Testing an XRootD cluster with SSI

1. Copy file to redirector node specifying storage path (/data/xrootdfs instead of /mnt/xrootd):

[email protected] # xrdcp /bin/sh root://<RDRNODE>:1094//data/xrootdfs/test1
[xrootd] Total 0.00 MB [================] 100.00 % [inf MB/s] 

1. To verify that SSI is working execute cns_ssi command on the redirector node:

[email protected] # cns_ssi list /data/inventory 
fermicloud054.fnal.gov incomplete inventory as of Mon Apr 11 17:28:11 2011 
[email protected] # cns_ssi updt /data/inventory 
cns_ssi: fermicloud054.fnal.gov inventory with 1 directory and 1 file updated with 0 errors. 
[email protected] # cns_ssi list /data/inventory 
fermicloud054.fnal.gov complete inventory as of Tue Apr 12 07:38:29 2011 /data/xrootdfs/test1 

Note: In this example, fermicloud53.fnal.gov is a redirector node and fermicloud054.fnal.gov is a data node.

(Optional) Enabling Xrootd over HTTP

XRootD can be accessed using the HTTP protocol. To do that:

1. Modify /etc/xrootd/xrootd-clustered.cfg and add the following lines.

      if exec xrootd
       xrd.protocol http:1094 libXrdHttp.so
       http.cadir /etc/grid-security/certificates
       http.cert /etc/grid-security/xrd/xrdcert.pem
       http.key /etc/grid-security/xrd/xrdkey.pem
       http.secxtractor /usr/lib64/libXrdLcmaps.so
       http.listingdeny yes
       http.staticpreload http://static/robots.txt /etc/xrootd/robots.txt
       http.desthttps yes
      fi
   

2. Add LCMAPS authorization configuration to /etc/xrootd/xrootd-clustered.cfg.

3. Create robots.txt. Add file /etc/xrootd/robots.txt with these contents:

      User-agent: *
      Disallow: /
   

4. Testing the configuration

   From the terminal, generate a proxy and attempt to use davix-get to copy from your XRootD host (the XRootD service needs running; see the services section). For example, if your server has a file named /store/user/test.root:

   davix-get https://<YOUR FQDN>:1094/store/user/test.root -E /mnt/xrootd/x509up_u`id -u` --capath /etc/grid-security/certificates
   

Note

For clients to successfully read from the regional redirector, HTTPS must be enabled for the data servers and the site-level redirector.

Warning

If you have u * in your Authfile, recall this provides an authorization to ALL users, including unauthenticated. This includes random web spiders!

(Optional) Enable HTTP based Writes

No changes to the HTTP module is needed to enable HTTP-based writes. The HTTP protocol uses the same authorization setup as the XRootD protocol. For example, you may need to provide a (all) style authorizations to allow users authorization to write. See the Authentication File section for more details.

(Optional) Enabling a FUSE mount

XRootD storage can be mounted as a standard POSIX filesystem via FUSE, providing users with a more familiar interface..

Modify /etc/fstab by adding the following entries:

....
xrootdfs                /mnt/xrootd              fuse    rdr=xroot://<REDIRECTOR FQDN>:1094/<PATH TO FILE>,uid=xrootd 0 0

Replace /mnt/xrootd with the path that you would like to access with. This should also match the GridFTP settings for the XROOTD_VMP local path. Create /mnt/xrootd directory. Make sure the xrootd user exists on the system. Once you are finished, you can mount it:

mount /mnt/xrootd

You should now be able to run UNIX commands such as ls /mnt/xrootd to see the contents of the XRootD server.

(Optional) Authorization

For information on how to configure xrootd-lcmaps authorization, please refer to the Configuring XRootD Authorization guide.

(Optional) Adding CMS TFC support to XRootD (CMS sites only)

For CMS users, there is a package available to integrate rule-based name lookup using a storage.xml file. If you are not setting up a CMS site, you can skip this section.

yum install --enablerepo=osg-contrib xrootd-cmstfc

You will need to add your storage.xml to /etc/xrootd/storage.xml and then add the following line to your XRootD configuration:

# Integrate with CMS TFC, placed in /etc/xrootd/storage.xml
oss.namelib /usr/lib64/libXrdCmsTfc.so file:/etc/xrootd/storage.xml?protocol=hadoop

Add the orange text only if you are running hadoop (see below).

See the CMS TWiki for more information:

• https://twiki.cern.ch/twiki/bin/view/Main/XrootdTfcChanges
• https://twiki.cern.ch/twiki/bin/view/Main/HdfsXrootdInstall

(Optional) Adding Hadoop support to XRootD

Hadoop File System (HDFS) based sites should utilize the xrootd-hdfs plugin to allow XRootD to access their storage:

1. Install the XRootD HDFS plugin package:

   [email protected] # yum install xrootd-hdfs
   

2. Add the following configuration to /etc/xrootd/xrootd-clustered.cfg:

   ofs.osslib /usr/lib64/libXrdHdfs.so
   

For more information, see the HDFS installation documents.

(Optional) Adding File Residency Manager (FRM) to an XRootd cluster

If you have a multi-tiered storage system (e.g. some data is stored on SSDs and some on disks or tapes), then install the File Residency Manager (FRM), so you can move data between tiers more easily. If you do not have a multi-tiered storage system, then you do not need FRM and you can skip this section.

The FRM deals with two major mechanisms:

• local disk
• remote servers

The description of fully functional multiple XRootD clusters is beyond the scope of this document. In order to have this fully functional system you will need a global redirector and at least one remote XRootD cluster from where files could be moved to the local cluster.

Below are the modifications you should make in order to enable FRM on your local cluster:

1. Make sure that FRM is enabled in /etc/sysconfig/xrootd on your data sever:

ROOTD_USER=xrootd 
XROOTD_GROUP=xrootd 
XROOTD_DEFAULT_OPTIONS="-l /var/log/xrootd/xrootd.log -c /etc/xrootd/xrootd-clustered.cfg" 
CMSD_DEFAULT_OPTIONS="-l /var/log/xrootd/cmsd.log -c /etc/xrootd/xrootd-clustered.cfg" 
FRMD_DEFAULT_OPTIONS="-l /var/log/xrootd/frmd.log -c /etc/xrootd/xrootd-clustered.cfg" 
XROOTD_INSTANCES="default" 
CMSD_INSTANCES="default" 
FRMD_INSTANCES="default"

1. Modify /etc/xrootd/xrootd-clustered.cfg on both nodes to specify options for frm_xfrd (File Transfer Daemon) and frm_purged (File Purging Daemon). For more information, you can visit the FRM Documentation
2. Start frm daemons on data server:

[email protected] # service frm_xfrd start
[email protected] # service frm_purged start

(Optional) Installing a GridFTP Server

The Globus GridFTP server can be installed alongside an XRootD storage element to provide GridFTP-based access to the storage.

See Also

OSG has extensive documentation on setting up a GridFTP server; this section is an abbreviated version documenting the special steps needed for XRootD integration. You may also find the following useful:

• Basic GridFTP Install. Additionally covers service planning topics.
• Load-balanced GridFTP Install. Covers the creation of a load-balanced GridFTP service using multiple servers.

Prior to following this installation guide, verify the host certificates and networking is configured correctly as in the basic GridFTP install.

Installation

GridFTP support for XRootD-based storage is provided by the osg-gridftp-xrootd meta-package:

[email protected] # yum install osg-gridftp-xrootd

Configuration

For information on how to configure authentication for your GridFTP installation, please refer to the configuring authentication section of the GridFTP guide.

Edit /etc/sysconfig/xrootd-dsi to set XROOTD_VMP to use your XRootD redirector.

export XROOTD_VMP="redirector:1094:/local_path=/remote_path"

Warning

The syntax of XROOTD_VMP is tricky; make sure to use the following guidance:

• Redirector: The hostname and domain of the local XRootD redirector server.
• local_path: The path exported by the GridFTP server.
• remote_path: The XRootD path that will be mounted at local_path.

When xrootd-dsi is enabled, GridFTP configuration changes should go into /etc/xrootd-dsi/gridftp-xrootd.conf, not /etc/gridftp.conf.
Sites should review any customizations made in the latter and copy them as necessary.

You can use the FUSE mount in order to test POSIX access to xrootd in the GridFTP server. You should be able to run Unix commands such as ls /mnt/xrootd and see the contents of the XRootD server.

For log / config file locations and system services to run, see the basic GridFTP install.

Using XRootD

Managing XRootD services

Start services on the redirector node before starting any services on the data nodes. If you installed only XRootD itself, you will only need to start the xrootd service. However, if you installed cluster management services, you will need to start cmsd as well.

The instructions for starting and stopping an XRootD service depend on whether the service is installed on an EL 6 or EL 7 machine, and whether you are using a standalone or clustered configuration.

On EL 6, which config to use is set in the file /etc/sysconfig/xrootd. For example, to have xrootd use the clustered config, you would have a line such as this:

XROOTD_DEFAULT_OPTIONS="-l /var/log/xrootd/xrootd.log -c /etc/xrootd/xrootd-clustered.cfg -k fifo"

To use the standalone config instead, you would use:

XROOTD_DEFAULT_OPTIONS="-l /var/log/xrootd/xrootd.log -c /etc/xrootd/xrootd-standalone.cfg -k fifo"

On EL 7, which config to use is determined by the service name given to systemctl. For example, to have xrootd use the clustered config, you would start up xrootd with this line:

[email protected] # systemctl start [email protected]

To use the standalone config instead, you would use:

[email protected] # systemctl start [email protected]

The services are:

Service	EL 6 service name	EL 7 service name
XRootD (standalone config)	xrootd	[email protected]
XRootD (clustered config)	xrootd	[email protected]
CMSD (clustered config)	cmsd	[email protected]

As a reminder, here are common service commands (all run as root):

To …	On EL 6, run the command…	On EL 7, run the command…
Start a service	service SERVICE-NAME start	systemctl start SERVICE-NAME
Stop a service	service SERVICE-NAME stop	systemctl stop SERVICE-NAME
Enable a service to start during boot	chkconfig SERVICE-NAME on	systemctl enable SERVICE-NAME
Disable a service from starting during boot	chkconfig SERVICE-NAME off	systemctl disable SERVICE-NAME

Getting Help

To get assistance. please use the Help Procedure page.

Reference

File locations

Service/Process	Configuration File	Description
xrootd	/etc/xrootd/xrootd-clustered.cfg	Main clustered mode XRootD configuration
	/etc/xrootd/auth_file	Authorized users file

Service/Process	Log File	Description
xrootd	/var/log/xrootd/xrootd.log	XRootD server daemon log
cmsd	/var/log/xrootd/cmsd.log	Cluster management log
cns	/var/log/xrootd/cns/xrootd.log	Server inventory (composite name space) log
frm_xfrd, frm_purged	/var/log/xrootd/frmd.log	File Residency Manager log

Links

• XRootD documentation