I've spent the past couple of days getting xgrid blast working in the BCRC. The documentation leaves a variety of things to be desired. I spent a few hours trying to figure out why the test workstation I had set up couldn't connect. Eventually, I figured out that I was having trouble because I had munged the preferences.plist file where it stores the LocalHostName. Xgrid doesn't like it when that field isn't defined. Once I got that straightened out, I was trying to figure out how to convert a set of fasta files into a searchable database. For that I did
cat *.fa | /usr/local/biotools/bin/formatdb -i stdin -oT -n database_name
Then, I couldn't get protein searches to work through Xgrid BLAST, even though I could by invoking it via the command line. It took me quite a while to figure out how to formulate a proper command line query too. Basically, I didn't realize that with blastall you have to refer not to the directory that contains the database, but to the pre-extension file names of the files in the directory, e.g.
/usr/local/biotools/bin/blastall -pblastp -d /Library/BLAST/database_name/database_name -i test.txt
Not exactly standard unix behavior. I finally got a command line query to work, but the same query wouldn't work through Xgrid BLAST. Eventually, I found that I hadn't noticed that a list of steps included a seventh step on the next page and I needed to add one more symbolic link so that Xgrid could find the BLOSUM62 scoring matrix. Now that it's working it looks like it will be great,though.
There's one persistent problem I haven't figured out how to fix yet: when machines that were connected run radmind and reboot, they don't seem to be able to rejoin the cluster. If I restart the cluster and reboot them, they can rejoin, so evidently something is getting whacked when they run radmind. Sigh... I can work around it -- sort of. Once I get them updated, I can just set radmind to not update and they'll be fine. But I still want to figure out what's getting munged and fix it.
StevenBrewer