Re: Chunks Down&/Out/Not?
Posted in 1993
Richard Ridley writes:
|> Running Informix On-line 4.00.UH1 on an HP 827 (HP-UX 8.02), when we
|> run tbmonitor it tells us that ALL chunks (including those of the
|> ROOTDBS) are Down. Also, running tbcheck -cD tells us that all
|> chunks are off-line.
|>
|> Now, the Online instance is up and running and we can still access
|> the data in our tables (Yes - the chunks are marked Y for mirroring
|> BUT have no mirror devices assigned to them)
|>
|> If we edit the tbconfig file and put a '0' in the 'mirror flag
|> position', Online refuses to bring the system up and says
|> 'Initialization failed - primary chunk down'.
|>
|> However (again), if we then change the mirror flag to 1 (still no
|> mirror device assigned), Informix lets us bring the instance on-line
|> and everything is hunky-dory EXCEPT it still tells us that ALL chunks
|> are down.
|>
|> Any ideas out there ? - we have been told to either dbexport or
|> tbunload and start from scratch, but there must be another
|> answer out there somewhere (unless of course there isn't....)
This is an example of a developer's nightmare: a bug that has to remain.
The bug, from the developer's perspective, is that setting MIRROR to 1
allows a system to come on-line when one or more chunks, with no
mirror available, are marked as down. Technically, this should not be
allowed; if no on-line chunk is available, the system should not be
allowed to come up (since there is no telling what sort of corruption
is present from whatever caused the chunk to go down). We have let
it remain because there is no other effective way of getting a system back
on-line quickly w/o it (users with multi-gig mission-critical systems do
not like hearing "restore your entire system from archive").
You should either restore from archive (after checking your disks for
bad sectors) or dbexport/dbimport. Without this, you run the risk of working
with a corrupted system.
OnLine 6.0 will feature smaller archive granularities. At that point,
this "workaround" will no longer work. For now, it will get you back
up and running.
Dave