Simulating corruption - Mailing list pgsql-general

From Jason J. Czerak
Subject Simulating corruption
Date
Msg-id XFMail.20000903183951.jason-czerak@jasnik.net
Whole thread Raw
List pgsql-general
Hey,

this is a little odd to be asking but... Anyone have any tools that woudl go in
and corrupt the database. Like just break things. And then have step by step
instructions on how to rebuild things and fix em up.  I know a restore from the
last backup tape would be best and easiest solution, but if rebuilding the DB
and just ridding the one little broken part is easier or better (say you wanna
not loose 4 hours of data by restoring but lose 10 minutes of data while
rebuilding)...


right now written a little perl script that is inserting 1,000,000 rows within
a test database. It's inserting numbers. 1,2,3,4,5,6...10321,10322...etc...
just a loop with num = num+1; insert num;   kinda thing.. so every line is
different.

And for thoses of you that are stats hungy..  the box is a K6 500 128 megs ram
and IDE drive.   redhat 6.9.5.  postgres 7.02.. perl DBI is the interface. X
is't being run on the machine.  but the Xterms are being displaed on my
workstation.   it's connecting to the DB server that's on it's own machine VIA
TCP sockets.

When I started it was going at 40 - 60% cpu.. 800 to 1Meg/sec disk access.... a
good 20 to 30 minutes into the loop. 90% to 100% CPU and disk access is at
400K/sec.... I think the CPU is the limiting factor on the bigger DB's and DISK
was the factor on smaller DB's... memeory isn't much of a problem since the
user+share BUFF and CACHE ranges havn't changed.  I wonder about turning off -F
like talked about before :)

ohh and the size of hte DB is now 14megs.




-- Jason




pgsql-general by date:

Previous
From: Tom Lane
Date:
Subject: Re: Indexes not working (bug in 7.0.2?)
Next
From: Waldemar Baraldi
Date:
Subject: How can I kow my permissions?