Conference rocks::dec_edi

3043.0. "PEDI problems...." by UTRTSC::SMEETS (Workgroup support) Thu Mar 06 1997 17:44

Hi,

Customer running DEC/EDI V3.1A Unix with Oracle RDB as database reported the
following problem.

Yesterday evening (18:33) TF's send via the PEDI gateway suddenly stayed in the
status SEND and Incoming TF's (00:38) didn't make it to the client application.

Looking in the decedi errorlog via cockpit or decedi_look showed the following
for every outgoing document since yesterday 18:33

Wed Mar  5 18:33:09 1997 PID =   810 NAME = PEDI Gateway
DECEDI__RIPDOCERR (e), error updating documents to PURGEABLE
DECEDI__DBERRLC (e), error loading cursor TLF_TRFSTAT
DECEDI__ERROR (e), unsuccessful completion
DECEDI__TRACEMOD (i), calling module = DECEDI__DB_ROLLBACK
DECEDI__SQLERROR (e), SQL error : SQLCODE = -1 : %RDB-F-IO_ERROR, input or
output error
-COSI-F-READERR, read error
-COSI-F-ECONNRESET, connection reset by peer
DECEDI__SQLOPER (e), SQL operation : open cursor
DECEDI__TRACEMOD (i), calling module = DB_LOAD_CURSOR
DECEDI__TRACEMOD (i), calling module = DECEDI__DB_TLF_OTRFSTAT
DECEDI__SQLERROR (e), SQL error : SQLCODE = -1 : %RDB-F-IO_ERROR, input or
output error
-COSI-W-ENDOFFILE, end of file

For the incoming document :

Thu Mar  6 00:38:17 1997 PID =   810 NAME = PEDI Gateway
DECEDI__EDISP2ENVIN (w), error disassembling envelope
DECEDI__EMAPORPARTIN (w), error mapping O/R name to connection id
DECEDI__ERRRCON (e), error reading connection record
DECEDI__ERROR (e), unsuccessful completion
DECEDI__COUNTRY (i), Country Code is 'NL'
DECEDI__AMDNAME (i), Administration Domain is '400NET'
DECEDI__PRDNAME (i), Private Domain is 'ABN-AMRO'
DECEDI__ORGNAME (i), Organization is 'EDI'
DECEDI__SURNAME (i), Surname is 'EB-CLIENT'
DECEDI__DBERRROT (e), error starting read only transaction
DECEDI__DBTABLE (i), database table = SCF
DECEDI__TRACEMOD (i), calling module = DECEDI__DB_ROLLBACK
DECEDI__SQLERROR (e), SQL error : SQLCODE = -1008 : %SQL-F-NOIMPTXN, no
implicit transaction to commit or rollback
DECEDI__TRACEMOD (i), calling module = DECEDI__DB_SCF_PEDI_START_TRANS_RO
DECEDI__SQLERROR (e), SQL error : SQLCODE = -1 : %RDB-F-IO_ERROR, input or
output error
-COSI-F-WRITERR, write error
-COSI-F-EPIPE, broken pipe

Stopping/starting decedi seem to solve the problems. After the restart of decedi
all outgoing/incoming traffic was handled without any problems.

Any idea what's the problem here. Is it a problem with the PEDI gateway or
an Oracle Rdb problem.

Thanks,

Martin

    >Any idea what's the problem here. Is it a problem with the PEDI gateway or
    >an Oracle Rdb problem.
    
    My initial reaction is it's an Oracle Rdb problem.
    
    Do they have `EDI Notification' turned on in the Connection Details?
    We may need to know this because updating the transmission file
    to PURGEABLE happens in different places in the code if `EDI
    Notification' is turned on.
    
    The second error indicates the PEDI gateway couldn't even start
    the read-only transaction to find the Connection Details record.
    Not much the PEDI gateway can do about that.

Hi David,

Last week, this customer performed an upgrade from V3.1 -> V3.1A and not 
everything went fine. So problems with the Oracle database could be possible.

Would it be a good idea to create a new database ?

Thanks,

Martin

Hi David,

It seems to be "normal" that the PEDI gateway deletes messages when disassembly
fails, but is it also normal that in such cases the PEDI gateways issues a
delivery notication to the partner.

To be more specific in the base note you can find that the imcoming message of
partner ABN-AMRO failed and was deleted but the partner received a delivery
notification.

I would expect that in these circumstances a non-delivery would be more 
appropriate.

Or is the MTA responsible for generating the delivery notification ?

Martin

    Martin,
    
    in cases of a useragent ( DECEDI is a Useragent ) the MTA is reponsible
    for the delivery notification.
    The UA successfully fetched the message.
    
    GJ
    

Hi David,

I've been onsite and indeed the problems were Rdb related, corruption of indexes
and other nasty problems.

Even resulting in strange things like the Cockpit summary view showed 20
available documents and the documents window showed no documents....

Martin