Installing an XRootD Storage Element¶
EL7 version compatibility
There is an incompatibility with EL7 < 7.5 due to an issue with the
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 the installation process, consider the following points:
- User IDs: If it does not exist already, the installation will create the Linux user ID
- Service certificate: The XRootD service uses a host certificate at
- 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 osg-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:
The eXtended Root Daemon controls file access and storage.
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
refer to the data node host.
These should be replaced with the fully-qualified domain name of the host in question.
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|
||Change to the hostname of the redirector|
||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
Further information can be found at https://xrootd.slac.stanford.edu/docs.html
Verifying the clustered config¶
cmsd on all nodes according to the instructions in the
Using XRootD section.
Verify that you can copy a file such as
/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
(Optional) Adding High Availability (HA) redirectors¶
It is possible to have an XRootD clustered setup with more than one redirector to ensure high availability service. To do this:
/etc/xrootd/xrootd-clustered.cfgon each data node follow the instructions in this section with:
set xrdr1 = <RDRNODE1> set xrdr2 = <RDRNODE2> all.manager $(xrdr1):3121 all.manager $(xrdr2):3121
- Advertise the
<RDRNODE>FQDN to users interacting with the XRootD cluster should be
(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
xrootddaemon 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:
- xrootd redirector
- xrootd - second instance that required for SSI
Host B is a data server that is running the following daemons:
- xrootd data server
- 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
We will be running two instances of XRootD on
/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.
root://<RDRNODE>:1094//data/xrootdfs/test1 will actually go to
Starting a second instance of XRootD¶
- Create a symlink pointing to
[email protected] # ln -s /etc/xrootd/xrootd-clustered.cfg /etc/xrootd/xrootd-cns.cfg
- Start an instance of the
cnsusing the syntax in the managing services section:
Testing an XRootD cluster with SSI¶
- 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]
- To verify that SSI is working execute
cns_ssicommand 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:
Configure LCMAPS authorization.
Uncomment the following line in
set EnableLcmaps = 1
Additionally, add the following line to the same file:
set EnableHttp = 1
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
davix-get https://<YOUR FQDN>:1094/store/user/test.root -E /mnt/xrootd/x509up_u`id -u` --capath /etc/grid-security/certificates
For clients to successfully read from the regional redirector, HTTPS must be enabled for the data servers and the site-level redirector.
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..
/etc/fstab by adding the following entries:
.... xrootdfs /mnt/xrootd fuse rdr=xroot://<REDIRECTOR FQDN>:1094/<PATH TO FILE>,uid=xrootd 0 0
/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.
/mnt/xrootd directory. Make sure the xrootd user exists on the system. Once you are finished, you can mount it:
You should now be able to run UNIX commands such as
ls /mnt/xrootd to see the contents of the XRootD server.
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
See this documentation.
(Optional) Adding Hadoop support to XRootD¶
For documentation on how to export your Hadoop storage using XRootD please see this documentation
(Optional) Adding Multi user support for an XRootd server¶
For documentation how to enable multi-user support using XRootD see this documentation.
(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:
- Make sure that FRM is enabled in
/etc/sysconfig/xrootdon 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"
/etc/xrootd/xrootd-clustered.cfgon 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
- Start frm daemons on data server:
(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.
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:
Prior to following this installation guide, verify the host certificates and networking is configured correctly as in the basic GridFTP install.
GridFTP support for XRootD-based storage is provided by the
[email protected] # yum install osg-gridftp-xrootd
For information on how to configure authentication for your GridFTP installation, please refer to the configuring authentication section of the GridFTP guide.
/etc/sysconfig/globus-gridftp-server to set
XROOTD_VMP to use your XRootD redirector.
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 full local path exported by the GridFTP server. For example
- remote_path: The XRootD path that will be mounted at local_path.
xrootd-dsi is enabled, GridFTP configuration changes should go into
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.
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
However, if you installed cluster management services, you will need to start
cmsd as well.
XRootD determines which configuration to use based on the service name specified by
For example, to have
xrootd use the clustered config, you would start up
xrootd with this line:
To use the standalone config instead, you would use:
The services are:
|Service||EL 6 service name||EL 7 service name|
|XRootD (standalone config)||
|XRootD (clustered config)||
|CMSD (clustered config)||
As a reminder, here are common service commands (all run as
|To …||On EL 6, run the command…||On EL 7, run the command…|
|Start a service||
|Stop a service||
|Enable a service to start during boot||
|Disable a service from starting during boot||
To get assistance. please use the Help Procedure page.
||Main clustered mode XRootD configuration|
||Authorized users file|
||XRootD server daemon log|
||Cluster management log|
||Server inventory (composite name space) log|
||File Residency Manager log|