thompson@ucselx.sdsu.edu (thompson s) (07/26/90)
Has anyone out there using INGRES 6.nn (preferably) on a VMS based
VAX experienced the following problem?
A user passes a query to the server (via ISQL) that is
non-trivial (eg. based on a complex view, etc..) and
decides that s/he doesn't really want to run the query
so s/he issues a <ctrl>C to abort the query...
The server apparently gets passed the "interrupt" and
puts the query's thread into a CS_DEAD_MASK state (this
was determined via IPM). Meanwhile the server goes into
a wild frenzy of disk thrashing (LOTS of DIO) and all other
threads on the server get marginal (if that) response.
The folks at INGRES seem to think that the server is doing
some sort of necessary clean-up, but this lasts for hours
and eventually we have to kill the server and start a new
one...
Help me if you can...
I'll even accept sym/empathy!
Bye Now,
Scott Thompson
San Diego State University
(619) 594-4276chesky@portia.Stanford.EDU (Snehylata Gupta) (07/26/90)
In article <1990Jul25.193815.7547@ucselx.sdsu.edu> thompson@ucselx.sdsu.edu (thompson s) writes: >Has anyone out there using INGRES 6.nn (preferably) on a VMS based >VAX experienced the following problem? > > A user passes a query to the server (via ISQL) that is Deleted stuff > The folks at INGRES seem to think that the server is doing > some sort of necessary clean-up, but this lasts for hours > and eventually we have to kill the server and start a new > one... > >Help me if you can... > >I'll even accept sym/empathy! > > Scott Thompson The Ingres people are right. We had the same problem running Ingres 6.3 on VMS 5.2 some one did a select * from table_1 where key like % and the table had 40,000 rows, after a while he did a cntrl - C. We watched that process for 6 hours through IPM and when we were about to kill the server it vanished. So in this case I agree with the INGRES folks. One thing that you can do is as follows : Raise the base priority of the process running that INGRES process. It should clean it up faster. Because I think that INGRES processes inherit the base priorities of the aprent process. Otherwise there is nothing you can do except shut down that server. You can bring up another server while the other one continues the rollback process. Sanjay