Re: hang/wait problem

Jared,

Like all my posts, this one is a long one too. Your question was easy but the answer was not. I remember things by just persistently "pinning " the required information in my instance's database buffer cache's KEEP segment. I then modify the behavior of my LRU algorithm by consuming required amounts of "Ginko Biloba" and "Ginseng", so that it stops aging blocks out of my memory. This feature however, is not supported and does not work as advertised on all "body platforms". My platform definitely supports it and it works...;-)

The KEEP, RECYCLE and DEFAULT segments of my database buffer cache have been specially re-configured per my requirements. After all, it is my system and I have full control over memory sizing, allocation and resource management, let alone the init.ora. It has some pretty cool dynamic memory expanding features....again not supported on all platforms.

I am currently running fine and have not yet experienced contention or saturation with my KEEP segment. Needless to say, I am yet to experience any coredumps. But I guess that will come with time and age...;-)

When I start experiencing ORA-600s and ORA-7445s related to access to my KEEP segment, and/or I start experiencing "memory leaks", I will listen to "music" hours on end, and hope I have some level of "Metal" support....crutches may be!.

By that time in my life, I may just have "Lead-level support". If and when they come on the line, they will recommend to patch the version of Oracle and the O-S on my node, to fix the problem, but that may not do it, because there may be no guarantee that when one shuts my instance down, that it may come back up....;-(

When all else fails, I will deploy my self-devised work around, by initiating an "online cache fusion" procedure. This procedure will synchronize my database buffer cache with the cache of the other instance in my "family cluster" which is configured in "Parallel Server".

This OPS configuration is one where the cluster configuration has a variety of different types of nodes of various ages and configurations. Despite the disparity in configuration and age, our systems will continuously "talk with each other", by "tnspinging" each other. Again, this configuration is not yet supported on all "platforms" and/or on all "family clusters".

The instance that I "fail-over" to, also will happen to be the one that was spawned by a "fork" system call from the PID of PMON on my Instance - It is my son's instance (Again forking complete Oracle Instances from single process PIDs is yet to be supported on all body platforms...;-)...You may have to wait for Oracle RDBMS version 10.3.4.7).

He is currently 4 years old now and would be at prime "system performance" levels at that time. When the time is right, his instance will then take over and someone can shut my instance down. Who knows it might just crash one day without warning!!!.  And that time automatic instance recovery may not revive the instance. It will then be time to then send a note out to "oracle-l_at_fatcity.com" and find out whether anyone else has experienced similar problems...;-)

So Jared, did I answer your question and also provide a sneak preview of my systems future high availability and scalability plans?....;-).

Cheers,

Gaja.

• Jared Still <jkstill_at_bcbso.com> wrote:
  >
  > Thanks Gaja.
  >
  > Too many features, too many versions, and I can't
  > keep it straight in my head as to which version
  > has what feature.
  >
  > How *do* you do it? :)
  >
  > Jared
  >
  >
  >
  > On Mon, 12 Jun 2000, Gaja Krishna Vaidyanatha wrote:
  >
  > > Jared my buddy,
  > >
  > > How goes it? Please allow me to make a small modification
  > to
  > > your response which says 7.x. The "warm start" feature of
  > > instance recovery was introduced in 7.3. The rollback is
  > done
  > > after the database is opened starting in 7.3.0 and above.
  > >
  > > Cheers,
  > >
  > > Gaja.
  > >
  > >
  > > --- Jared Still <jkstill_at_bcbso.com> wrote:
  > > >
  > > > If the session is waiting on rollback, you will have to
  > > > wait it out anyway after bouncing the DB in ORacle 7.x, as
  > > > the rollback must complete before the DB is opened.
  > > >
  > > > Jared
  > > >
  

...stuff deleted....

Gaja Krishna Vaidyanatha   | 3460 West Bayshore Road,
Manager - Integration      | Palo Alto, CA 94303
& Consulting Services      | gaja_at_brio.com
Global Alliances           | (650)-565-4442
Brio Technology            | www.brio.com 

"Opinions and views expressed are my own and not of Brio Technology"