collect

[Thomas J. Harrold]

I'm trying to use collect (playback mode) to help me figure out when a filesystem (domain) ran out of space, and then try to determine what was running at that time.

Man page for cfilt is useless, but I've tried:
collect -p | cfilt -u 'file:free' | sort -k3 | more

This shows me the lowest amount of free space, aggregated for all filesystems, but doesn't really help me figure out when free space went to 0. Is there a way to do this, without having to write a complex shell/perl script, to show me what I'm looking for?

Thanks,

-tjh

[Mark Poeschl_2]

Presumably it's a specific filesystem you're after... If so:

# collect -s f -p | cfilt file:filesystem=:free | more

should spit out the time tag, sample #, and free space for each sample in the collect data file. Just wait for the third column to get to zero and note the time tag. Then you can go back into the collect file for that time and see what processes are running.

[Venkatesh BL]

You could also use collgui (the graphic utility for viewing collect data) to check out the flow of things.

[Thomas J. Harrold]

Took me a while to figure out the format of the filesystem. (ie root_domain#root)

Is there a document that contains all of the cfilt options? I was unable to find anything official...

-tjh

[Mark Poeschl_2]

All the different data fields are documented in the man page for collect.

[Venkatesh BL]

...and the manpage of cfilt contains some examples of the expressions that can be used.

The main consumer of cfilt is the Collgui tool, which uses it to retrieve specific data from collect datafiles.