[developers] Need help with the feature caching step of the parse ranking code in the logon tree

Bill McNeill (UW) billmcn at u.washington.edu
Thu Dec 11 01:18:12 CET 2008

I am trying to set up my system to do parse ranking experiments.  I have
checked out the logon tree from http://svn.emmtee.net/extras/handon.  I
tried to run scripts in the logon/ling/redwoods directory as per
instructions <http://wiki.delph-in.net/moin/LogonModeling> on the Wiki;
however, the feature caching step fails.

Specifically, I ran the command "load --binary fc.lisp" in the
logon/lingo/redwoods directory which collects features from the jhpstg
corpus.  It ran for 2 hours and 45 minutes and appeared to be working
correctly, but then it failed with the following error.

[17:45:41] cache-features(): item # 31212: 7 events;
[17:45:41] cache-features(): item # 31213: 7 events;
[17:45:43] cache-features(): item # 31243: 294 events;
[17:45:43] cache-features(): item # 31251: 4 events;
[17:45:43] cache-features(): item # 31252: 3 events;
[17:45:43] cache-features(): item # 31253: 3 events;
[17:46:02] operate-on-profiles(): running `jh3' [31261 - 31511|.
[17:46:13] gc-after-hook(): local; new: 1677889568; old: 0; efficiency: 80.
Error: open-fc(): error -1 for

Restart actions (select using :continue):
 0: retry the load of fc.lisp
 1: skip loading fc.lisp
 2: recompile /home/billmcn/logon/lingo/redwoods/fc.lisp
 3: Return to Top Level (an "abort" restart).
 4: Abort entirely from this (lisp) process.
[1] TSNLP(4): :pop
Do you really want to kill process "Initial Lisp Listener" [n]?

The file /home/billmcn/logon/lingo/redwoods/tsdb/home/jh3/fc.bdb exists.  At
the time of the error the file fc.bdb exists and is 2959577088 bytes long.
There was a message in STDERR that read

db_open(): error `DB_RUNRECOVERY: Fatal error, run database recovery'

I tried running it again and saw a similar error, except this time in

I'm stuck and need help getting past this step.  Any suggestions?  Here are
my specific questions.


   1. Any idea what the problem is?
   2. Can I just rerun load --binary fc.lisp, or should I do some cleanup
   3. How do I run database recovery as directed by the error message in
   4. Is there some way I can specify a very small treebank for the system
   to work with?  Something that could finish this step in a few minutes rather
   than a few hours.  It would be okay if it was too small to get reliable
   statistics out of, since for starters I'm just verifying if the setup works.
   5. Is there some way I can manually run the individual jh1, jh2, jh3,
   etc. steps?  (It seems like they shouldn't depend on each other.)

This is being run on the patas system at the University of Washington, which
is Red Hat Linux.

Bill McNeill
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.delph-in.net/archives/developers/attachments/20081210/4ef3731a/attachment.html>

More information about the developers mailing list