Re: [Info-Ingres] infodb iidbdb freezes

Hi Marty,

I've experienced infodb freeze in a few scenarios. I'm not 100% sure of
the order of events because I would be called in at the point of freeze.


Run infodb | more and pause the output. Then run infodb on another
session.
I think this is the same for both windows and unix. It has been around
since ingres2.6


Nightly checkpoint stalls waiting for open transactions to complete.
(forgot the timeout param)
The following night the checkpoint locks up
DB operations such as createdb or destroydb lock up
infodb locks up.


Destroydb stalls whilst someone is connected to the db.
Even after disconnect or terminate of the sessions doesnt fix it.
Need to manually kill the destroydb process.
infodb at some stage locked up.

good luck



________________________________

From: info-ingres-bounces@kettleriverconsulting.com
[mailto:info-ingres-bounces@kettleriverconsulting.com] On Behalf Of
Martin Bowes
Sent: Monday, 8 February 2010 11:47 PM
To: Ingres and related product discussion forum
Subject: [Info-Ingres] infodb iidbdb freezes



Hi All,



I have a windows installation (I'm sorry...) II 9.2.0 (int.w32/143) +
p13595.



The system just bevaed in a very strange manner over the weekend.



An attempt to do infodb iidbdb would freeze. At the time a backup of
iidbdb had occurred but would not conclude. At all times the server was
accessible and you could connect to iidbdb and run queries.



I was forced to user taskkill to remove infodb requests and the
checkpoint process. After I managed to stop ingres the
files/memory/lglkdata.mem file was not deleted.



I restarted ingres and reran the backup and once more it froze on iidbdb
having completed the creation of the checkpoint. Ditto infodb iidbdb
would freeze. And again after the shutdown the lglkdata.mem file
remained.



I restarted the installation and altered my backup script to not use
'#m2'. It now appears to be ok in backups...but there have been several
backups with this flag in effect, its only the last weekend where there
appears to have been a problem.



Has anyone seen anything like this or have any suggestions to offer?



Martin Bowes