Re: Oracle lock timeouts

In article <rqEYsApKBh107h_at_uidixon.uucp> uidixon!idixon_at_infocom.co.uk (Ian Dixon) writes:

>In <mdayton.21_at_stpaul.ncr.com> mdayton_at_stpaul.ncr.com (Mike Dayton) writes:
>>In article <23m3vk$ijg_at_pandora.sdsu.edu> oliver_at_io.nosc.mil (George Oliver) writes:
 

>[Lots of stuff deleted]
 

>>The work around of killing the shadow process works only if you can
>>identify the process to kill. In version 6 you could not make the
>>association between an Oracle PID and a System PID. [ stuff deleted ]
 

>I seem to have to do this every day so it can be done. Go into SQL*DBA
>and use 'monitor user' to find the Oracle user of the PC client that has
>crashed. Part of this shows the system pid which can then be killed. OK,
>its a bit awkward and it relies on the user having only one login at the
>time but it works. This is under Oracle v6 on a SCO Unix server with
>PC clients connect via FTPs PC/TCP.

OK, I should be more careful when making generalizations. I am using Wollongong's Pathway Access TPC/IP on the PC clients. For Servers I have NCR Towers running RDBMS 6.031 and ATT Starserver E's running 6.036.

When I use 'monitor user' I see a real system PID for the background processes, clients running on the database server and clients running over SQL*Net where the remote machine is a UNIX box. However, for all PC clients running SQL*Net over Pathway Access the System PID is always of the form "xxxx:01" where xxxx is not a process which is running on the database server.

Does anyone know what the "xxxx:01" is? Could the TCP/IP on the PC be the cause of this? Is it possible to map it to a *real* system PID?

Thanks.

...............................................................................

Mike Dayton                                     | AT&T / NCR
Software Services Automation                    | 2700 Snelling Avenue North
email: Mike.Dayton_at_StPaul.NCR.COM               | St. Paul, MN. 55113