[Orca-users] Re: rebuild rrd files from percol files

[jon.tankersley at sabre.com]

I've got some more 'input' for you guys on this one.

I've been seeing this too.  Three directories with percol files in  
them:  XXX/percol...2001-01-20.gz through 30.gz.  YYY with 
2001-01-05.gz through 30.gz, ZZZ with 01-05 through 01-30.

Destroying the RRD and html and png and ... and running orca again 
gives valid 'daily' plots, but the 'weekly' has 01-21 through 01-24 
missing from EACH system, despite the existance of the percol files.

Some interesting issues noticed with the -v flag for orca.

01-29 01-20 01-21 01-27 01-22 ... are the order of the percol files 
processed.

Some data lines in the percol files don't have the right number of 
columns.  This should be eliminating that data point, not the entire 
file.  I'm also not sure WHY this is happening, but then I'm testing 
some hacks to diskinfo.se in the SEToolkit.

The data was gathered out of order (ftp problems).  Rebuilding the 
directories XXX,YYY,ZZZ with the files in the correct order didn't do 
any good.  Same 'parse' order as above.

I've started a rebuild copying one file at a time into the XXX,YYY,ZZZ 
directory and running orca -o.  It isn't done yet.

>From the code, it looks like the names are stored in a hash instead of 
an array or some such.  This seems to be overriding the date order.