-
If not yet done, copy the distribution network_drive
directory to the H: network drive.
-
Edit
H:\CBSU\clusters. This directory contains text files, each file
corresponds to a cluster and lists all its nodes.
-
Copy your BLAST databases to
the appropriate H:\CBSU subdirectories (optional). You can
download databases later, see documentation for more details.
-
Edit H:\CBSU\updates\tasks
and H:\CBSU\updates\tasks.repeat. These files define scripts that
are executed on each node before BioHPC calculations start. The
first script runs only once on a node, the other each time when a
job starts.
-
Install SQLExpress Advanced,
choose mixed mode of authentication (Windows and server/local) for
database access.
-
Copy CBSUJOBS database
from installation to the local data directory of your SQL server. Attach
this database.
-
Create an account for
database access (recommended name biohpcsql) with local SQL server authorization, set password
(later needed in setup.aspx).
-
Allow biohpcsql to write data
to database and execute stored procedures newdb, newjob, newrtjob.
-
Install Perl.
-
Install Visual Studio 2008
(optional). You will need Visual Studio if you want to recompile
programs or interface.
-
Install Windows SDK and .NET 3.5
runtime
-
On the web server Install client
for Microsoft Windows Compute Cluster Server 2003 and
client for Microsoft Windows HPC Server 2008. You need
to install at least the one that corresponds to the cluster
scheduler you will be using, you may install both.
-
Set up IIS virtual directory
(recommended name BioHPC) for BioHPC with ASP.NET and a special Application Pool
running as IDU
-
Enable ASP .NET Web Extension
in IIS
-
Sometimes there are problems deploying ASP.NET on
the server when it is the first ASP.NET site on the server. You may need
CleanIISScriptMaps.exe if there are problems with registering BioHPC
with .NET 2. Check .NET registration status in "Properties" of the
virtual directory.
-
Enable SQL Express TCP/IP
channel (if using SQL Express)
-
Enable mixed authorization
modes in SQL (local and Windows, parameter LoginMode
2). You will need to do it only if you forgot to select this
option during installation. You can do it via SQL Server
configuration utility, or directly in the registry. After the
changes are done SQL services have to be restarted, especially SQL
Agent.
-
Set up virtual directories
JOBS (pointing to H:\CBSU\JOBS) and CBSUFTP (optional), setup them to display various file types
other than html or aspx as text (MIME: .*
text/plain)
-
Add IDU to iis_wpg group on
web server (Windows Server 2003). Add IDU to iis_usrs group on web
server (Windows Server 2008).
-
Make web.config file fully readable/writable
by IDU
-
Make sure all directories in
BioHPC web site are
accessible to IDU
-
Create file H:\CBSU\BioHPC_url
readable/writable to IDU with full INTERNAL (from CCS/HPC cluster) url to BioHPC
web server.
-
Create echeck.log with full
control for IDU in BioHPC web directory.
-
Make sure local drives T: from the nodes are
exported as
t-drive with full control access for IDU
-
Make messages subdirectory in BioHPC web directory
with full
control access for IDU
-
Open file web.config (in a text editor like
notepad) from directory BioHPC and change
-
<add key="WebDir" value="H:\Inetpub\wwwroot\BioHPC\"
/>
such that value points to the directory where web server files
for BioHPC are installed
-
<add key="HDriveUnc"
value="\\fileserver\share" />
such that value points to a share on the fileserver which is
supposed to host BioHPC network drives (H:\CBSU). If there is a
local H: drive and BioHPC files are installed in H:\CBSU
directory do nothing, server always checks this directory
existence first.
-
Leave other tags unchanged, they will be
modified later.
-
On the web server, open http://localhost/BioHPC/setup.aspx
(use appropriate path to BioHPC virtual directory) and set local parameters. This file won't open outside the server.
-
If using SQL Express locally, the server name is
server\SQLEXPRESS
-
Login as admin@biohpc with initial password
BioHPC4All . Change initial administrator id to a valid e-mail address and change
the password.
-
Set up ftp server for BioHPC (optional). Create
JOBS and PBLAST directories at ftproot, writable to IDU locally.
Create virtual directories JOBS and PBLAST readable to IDU. Export
JOBS and PBLAST as UNC from the server, full access for IDU. If
the FtpJobsUnc parameter in setup.aspx is empty, the it is assumed
that the ftp server is not installed.
-
Modify acknowledgemnt.htm
-
Recompile ptools if necessary (present in H:\CBSU\src).
Update binaries at H:\CBSU\ptools\mpims and H:\CBSU\bin_x64 using
files from ptools\debug.
-
Set up automated tasks in Task Scheduler.
-
JSDLupdate. This service is required for JSDL
clusters. Add a task (as IDU) for web_dir\JSDLupdate\bin\Debug\JSDLupdate.exe,
running every 3 minutes, with no timeout (i.e. 72 hrs timeout).
This task may take long when downloading files from JSDL
cluster, but new task won't start until the previous one
finishes.
-
StarDaemon. This service will update job
information in BioHPC database, send required e-mail to the
users without need of administrator action. Add a task (as IDU)
for web_dir\StarDaemon\bin\Debug\StarDaemon.exe, running every
30 minutes, with no timeout.
-
If you plan to use HPC Basic Profile in order to access remote
clusters you will need to install a supplementary web service
providing information about available resources that is missing in
the HPC Basic Profile implementation. Otherwise BioHPC will not know
how many nodes/cores are available at the moment and it will not be
able to submit jobs based on resource availability. Download the web
service files and unzip them into a directory inside your web
server. Create an active directory for the service and run it in
the special Application Pool running as IDU.