Bingo index creation error (postgresql 9.1 on Windows 7, 64-bit)

[M Miller]

Hello,

I’m running into an issue trying to build a bingo structure table for sss/similarity searching.  (Bingo 1.7 beta, 64-bit Postgresql 9.1 on Windows 7, 64-bit.)

The table consists of 4 fields: 2 ints, on text field and one float. (The float field is not currently populated)

After creating the table, I loaded an SD file using bingo.ImportSDF()

And saw no errors.

 

CheckMolecule revealed some issues, so I build a clean version of the table using

 

create table structs_clean_mols as select * from bingo.structs where bingo.CheckMolecule(molfile) is null;

create index structs_clean_mols_index on structs_clean_mols  using bingo_idx (molfile bingo.molecule).

 

The ‘create index’ cranks along for a a few minutes, last new information:

NOTICE:  processed 3000 structures

Then, postgres displays the error message

server closed the connection unexpectedly

         This probably means the server terminated abnormally

         before or while processing the request.

The connection to the server was lost. Attempting reset: Failed.

In the postgres log, I see

2012-03-12 21:26:54 EDT LOG:  server process (PID 6052) was terminated by exception 0xC0000374

2012-03-12 21:26:54 EDT HINT:  See C include file "ntstatus.h" for a description of the hexadecimal value.

 

Googling, I found “0xC0000374 A heap has been corrupted.”

 

The postgres windows service stopped. Tried restarting and building the index again.

Similar message.

 

Any ideas?

[Michael Kvyatkovskiy]

Hello,

Seems like it is some problems in Bingo memory management. I have
reproduced a similiar error, and we will try to fix it soon.

Best regards,
Mikhail.

[M Miller]

Update: I tried setting up Bingo for Oracle to get familiar with Bingo operations while waiting for the fix for Postgresql.

I got as far as index creation and got a different error, pasted below.

Is this a manifestation of the same issue?


Error starting at line 1 in command:
CREATE INDEX mols_clean_index ON mols_clean (molstruct) INDEXTYPE IS Bingo.MoleculeIndex
Error at Command Line:1 Column:13
Error report:
SQL Error: ORA-29855: error occurred in the execution of ODCIINDEXCREATE routine
ORA-20352: Error: Unknown exception
ORA-06512: at "BINGO.MANGOCREATEINDEX", line 1
ORA-06512: at "BINGO.MANGOINDEX", line 13
29855. 00000 -  "error occurred in the execution of ODCIINDEXCREATE routine"
*Cause:    Failed to successfully execute the ODCIIndexCreate routine.
*Action:   Check to see if the routine has been coded correctly.

[Mihail Kvyatkovskiy]

Hello,

Such situation can be caused by lack of Bingo privileges on the molecule table. To create index you need to grant select privileges to Bingo using a command like this:

grant select on mols_clean to bingo;

Best regards,
Mikhail Kvyatkovskiy.

[M Miller]

Thank you, Mikhail, for the prompt reply!!

I'm running the index creation command from the same account (bingo) that owns the table.
Is there another privilege to check?

Regards,
Mitch

[Mihail Kvyatkovskiy]

Hello Mitch,

 

Due to the fact that you manage the tables using Bingo account, the problem is not with privileges. The information from Bingo log file may help to determine what is causing index creation failure, so may I ask you to paste data from bingo.log file here? On Linux it is located in /tmp/bingo.log. For getting clear data you may execute following commands:

 

$ mv /tmp/bingo.log /tmp/bingo.log.backup

$ sqlplus bingo/password@sid # Run index creation proceude

$ cat /tmp/bingo.log

 

Best regards,

Mikhail Kvyatkovskiy.

[M Miller]

Thanks, Mikhail!

OK; I'm on Windows so the actual commands are different but the log file is in the same place.

I see 12k lines like
preparing molecule #12809 with rowid AAASopAAAAAAFXFAAX

and then:
Error: Unknown exception

Is there any way I can obtain more detailed information?

Best.
Mitch

On Monday, March 12, 2012 9:52:01 PM UTC-4, M Miller wrote:

[Mihail Kvyatkovskiy]

Hello Mitch,

Yes, you can provide us the molecule that may cause the error. It can be done using such query:

select * from mols_clean where rowid = 'AAASopAAAAAAFXFAAX';

We will try to reproduce your problem using this molecule.

In general it looks like that problem is caused by unexpected exception, like division by zero or some memory management errors. 

As for memory management, could you also provide any information about extproc and oracle processes memory consumption before error appearance? Is it growing or not? Is there enough free memory?

Best regards,

Mikhail Kvyatkovskiy.

[M Miller]

FYI, I've e-mailed the molecule to you, Mikhail.

Please let me know what you find.

Mitch

On Monday, March 12, 2012 9:52:01 PM UTC-4, M Miller wrote:

On Monday, March 12, 2012 9:52:01 PM UTC-4, M Miller wrote: