Hi Peter,
FWIW, the fsstat code calls the following for each kstat
at each interval:
- kstat_chain_update()
- kstat_lookup()
- kstat_read()
If any of those fail, then we take it out of the list
of "entities" (fstypes or mount-points) that we're
watching.
While on the subject of kstats: PLEASE be aware that
the stability classification of the kstat name is still
private and is subject to change. This change may be
necessary to accomodate additional feature support.
Rich
Peter Tribble wrote:
> On Fri, 2006-03-24 at 23:46, Tom Erickson wrote:
...
>>I was wondering what a square in a data cell means. I attached a
>>screenshot in case you don't see these on your system.
>
>
> I don't see those. My first guess is that the filesystem has
> been unmounted. I didn't check for that in jfsstat (it never
> happens on my test systems unless I explicitly do a mount
> or unmount so didn't get tested). Error/sanity checking wasn't
> as complete as it should be :-( The next version will catch
> this and remove the entry from the list. (The version after
> might even spot new entries and add them on the fly.)
>
> This actually revealed sloppy coding in the C part - I was
> relying on kstat_lookup failing to tell me if a kstat had
> been removed, and it wasn't (probably because I hadn't called
> kstat_chain_update).
>
