Installing an Open Science Pool Access Point¶
This document explains how to add a path for user jobs to flow from your local site out to the OSG, which in most cases means that the jobs will have far more resources available to run on than locally. If your local batch system frequently has many jobs waiting to run for a long time, you do not have a local batch system, or if you simply want to provide a local entry point for OSG-bound jobs, adding a path to OSG may result in less waiting for your users.
Note that if you do not have a local batch system, consider having your users use OSG Connect, which will require less infrastructure work at your site.
Flocking to the OSG requires some modification to user workflows. After installation, see the usage section for instructions on what your users will need to do.
Every batch computing system has one or more entry points that users log on to and use to hand over their computing work to the batch system for completion. For the HTCondor batch system, we say that users log on to a access point (i.e., submit node, submit host) to submit their jobs to HTCondor, where the jobs wait ("are queued") until computing resources are available to run them. In a purely local HTCondor system, there are one to a few access points and many computing resources.
An HTCondor access point can also be configured to forward excess jobs to an OSG-managed pool. This process is called flocking. If you already have an HTCondor pool, we recommend that you install this software on top of one of your existing HTCondor access points. This approach allows a user to submit locally and have their jobs run locally or, if the user chooses and if local resources are unavailable, have their jobs automatically flock to OSG. If you do not have an HTCondor batch system, following these instructions will install the HTCondor submit service and configure it only to forward jobs to the OSG. In other words, you do not need a whole HTCondor batch system just to have a local OSG access point.
Because jobs running in OSG run in a completely different environment than your local cluster, and require additional considerations for data movement, you or your campus will need to provide additional support for your users to enable them to properly use OSG. Send email to email@example.com saying that you would like to flock jobs to OSG, and we will consult with you about the proper solution for your site and what needs to be set up at both the OSG and at your site.
Also consider the following technical requirements:
- Operating system: A RHEL 7 compatible operating system.
- User IDs: If it does not exist already, the installation will create the Linux user ID
- Host certificate: Recommended for authentication to OSG infrastructure. See our documentation for instructions on how to request and install host certificates.
- Inbound TCP port 9618 must be open.
- The access point must have a public IP address with both forward and reverse DNS configured.
As with all OSG software installations, there are some one-time steps to prepare in advance:
Scheduling a Planning Consultation¶
Before participating in the OSG, either as a computational resource contributor or consumer, we ask that you contact us to set up a consultation. During this consultation, OSG staff will introduce you and your team to the OSG and develop a plan to meet your resource contribution and/or research goals.
Register your access point in OSG Topology¶
To be part of OSG, your access point should be registered with the OSG.
You will need information like the hostname, and the administrative and security contacts.
Follow the general registration instructions.
For historical reasons, the service type is
Choose your authentication method¶
There are two options for authentication of your access point to the OSG: GSI and pool password. Of these, GSI is the recommended method, but it will require obtaining a host certificate for your access point. See our documentation for instructions on how to request and install host certificates.
If you are unable to obtain a host certificate, use the "pool password" authentication method, described in the Configuring Authentication section.
Request to be allowed to flock to OSG¶
OSG staff will need to add your access point to the list of hosts that flocked jobs are accepted from. Send email to firstname.lastname@example.org with the hostname of your access point and request to be added to the list.
Installing Required Software¶
Flocking requires HTCondor software as well as software for reporting to the OSG accounting system. OSG provides a convenience RPM that installs all required packages. To install, run:
[email protected] # yum install osg-flock
Configuring Reporting via Gratia¶
Reporting to the OSG accounting system is done using the Gratia service, which consists of multiple probes.
HTCondor uses the "condor" probe, which is configured in
we provide a recommended configuration for flocking.
Copy over the recommended probe configuration:
[email protected] # cp /etc/gratia/condor/ProbeConfig-flocking /etc/gratia/condor/ProbeConfig
Fill in the value for
ProbeNamewith the hostname of your access point, with the following format:
Fill in the value for
SiteNamewith the Resource Name you registered in Topology (see instructions above). For example:
Enable the Gratia Probe.
Under the section
Title2make sure to set the following (if not already there):
MapUnknownToGroup="1" MapGroupToRole="1" VOOverride="OSG"
If you cannot obtain a host certificate, then configure pool password authentication via the following instructions:
Send email to email@example.com requesting the pool password file for flocking.
Move the pool password file you have received to
/etc/condor/pool_password; set the ownership of that file to
condor:condorand the permissions to
Create a configration file named
/etc/condor/config.d/90-flock-pool-password.confwith the following contents:
#-- Flock to OSG using pool password SEC_PASSWORD_FILE = /etc/condor/pool_password SEC_DEFAULT_AUTHENTICATION_METHODS = FS,PASSWORD
In addition to the HTCondor service itself, there are a number of supporting services in your installation. The specific services are:
||See CA documentation for more info|
The following table gives the commands needed to start, stop, enable, and disable a service:
|To...||Run the command...|
|Start a service||
|Stop a service||
|Enable a service to start on boot||
|Disable a service from starting on boot||
Running jobs in OSG¶
If your users are accustomed to running jobs locally, they may encounter some significant differences when running jobs in OSG. Users should be aware that OSG jobs are distributed across multiple institutions across a large geographical area. Each institution will have its own policy about the kinds of jobs that are allowed to run, and data transfer may be more complicated. The OSG Helpdesk Solutions page has information about what users should know; the Choosing Resources for Jobs and Data Management sections are particularly relevant.
Specifying a project¶
OSG will only run jobs that have a registered project associated with them. Users must follow the instructions for starting a project in OSG-Connect to register a project.
A project is associated with a job by adding a ProjectName line to the user's submit file. For example:
+ProjectName = "My_Project"
The double quotes are necessary.
If you need help with setup or troubleshooting, see our help procedure.