Re: [Gmod-ajax] AJAX Gbrowse - problem rendering tiles for EST reads

[Mitch S. <mit...@be...>]

Adam Davis wrote:
> The "wo oo" part was me inserting prints into generate-tiles.pl  to 
> find the point at which it hangs.  Here is the code from that file, 
> starting at line 255:
> The print statement I inserted actually gets cut off halfway through, 
> which is odd.  It always cuts off in the same place too.
>
> print
>    " There are $num_tracks tracks, $num_zooms zoom levels total\n",
>    
> "-------------------------------------------------------------------------\n"; 
>
> print " wo\noo\noo"; 
> <---------------------------------------------------------------------------------------Here 
> is the print I inserted

Thanks for the detailed reply.

Hmmm, I probably need to spend some time reading the code and thinking,
but off the top of my head I have some information that might help.  The
wo\noo\noo gets cut off before the last oo because the output is line
buffered; your last oo does get output to some buffer, but the program
gets stuck before the oo makes it any further than that.

So it might seem like the program is getting stuck in the middle of your
print statement, but it's actually further down than that.  You could
try moving the print further down, and/or you could also try running
generate-tiles with strace:
strace ./generate-tiles.pl <all your generate-tiles.pl arguments>
I'm assuming you're using linux.  Strace will probably generate a lot of
output, and we'd be interested in the last few screenfuls of it.

Whenever something hangs on linux and it's not using any CPU (it isn't,
right?) it's usually waiting on some kind of IO; best guess is the
database.  We've tried to use Bio::DB similarly to the way regular
GBrowse does, so that we'll be compatible with existing GBrowse setups,
but I think we have to do some things differently, and this is probably
due to one of those.

One of the things we do differently from regular GBrowse is that we deal
with an entire landmark (chromosome/scaffold/contig/whatever) of data at
a time (to do layout).  So the hang may just be the database churning,
trying to gather all the data that we're asking for.  How many reads are
on scaffold_1?

I expected some difficulty scaling with some plant genomes, which I've
heard are huge, but I thought the individual chromosomes weren't much
bigger than (say) human, so I was hoping we'd be okay.  But if my guess
is right then we need to do some more scaling work.

Mitch