Hello !

I just found an article published in a Linux magazine in Brazil about some
tests with IB running on Linux related to IB crash recovery system.

The article (in portuguese) can be found at http://www.revistadolinux.com.br/artigos/database/20010117084829.php3

Here is a resume of what they did. I would like to hear some comments about
their conclusions :

They did basicly 2 kinds of crash tests :

1) Killing the main process (and its subprocess) using the linux KILL
command during some operations (insert,update,etc.)

2) RESETing the hardware during some operations (insert,update,etc.)

In the first case, IB recovery system was 100% in almost all cases . Only
in 1 situation an error was found : While updating 300,000 records with
forced write ON, after killing the process and trying to connect to the
database again, the connection was accepted and NO error message was
reported, but when the user tried to access the table that was being
updated the system freezed.

In the second case, IB returned errors when trying to connect to the
database :

üdatabase file appear corrupt ( )
wrong page type
page "xxxxx" is of wrong type (expected 7, found 5)

In some of them, the database could be connected but the table was
inaccessible :

Internal GDS software consistency check (cannot find record back version (291)) (3 ocorrências)
§Internal GDS software consistency check (wrong record length (183))
§Internal GDS software consistency check (applied diferences will not fit in record (177))
§database file appear corrupt ( )
wrong page type
page "xxxxx" is of wrong type (expected 5, found 4)



[]s

Carlos
WarmBoot Informatica - http://www.warmboot.com.br
Interbase-BR - http://www.interbase-br.com