dtrace discuss - Nov 2008 - displaying queues of running cron

Prior to Solaris 10, one could easily enough display the contents of the struct
queue array
qt (see cron.c) using mdb; that shows, for each queue (all 26 possible, not just
the
standard a b c), the values in queuedefs(4), plus the number of jobs currently
running in
that queue.

But from Solaris 10 on, something must be different about the way cron is
linked; mdb
can no longer see that symbol.  (unless there''s some trick I''m
missing)

With Solaris >= 10, it might be nicer if there were dtrace probes to retrieve
that info...
something like a proper interface, as it were.  And a few more probes (job
start, job end (with return code), errors affecting job scheduling, etc) perhaps
might be nice too, enough
to effectively monitor cron''s behavior.  Oh, and one other suggested
probe: a way to get
cron''s idea of the timezone it is running in (although pargs could
retrieve its TZ value,
it would be better of all the interesting stuff could be done one way).

Anyone looking into adding probes to cron yet?
-- 
This message posted from opensolaris.org

> Prior to Solaris 10, one could easily enough display the contents of  
> the struct queue array
> qt (see cron.c) using mdb; that shows, for each queue (all 26  
> possible, not just the
> standard a b c), the values in queuedefs(4), plus the number of jobs  
> currently running in
> that queue.
>
> But from Solaris 10 on, something must be different about the way  
> cron is linked; mdb
> can no longer see that symbol.  (unless there''s some trick
I''m
> missing)
I''m not sure what you and I are doing differently, but I can see that  
symbol:

 > qt::nm
Value      Size       Type  Bind  Other Shndx    Name
0x0806dc48|0x000001a0|OBJT |LOCL |0x0  |23      |cron`qt
> With Solaris >= 10, it might be nicer if there were dtrace probes to  
> retrieve that info...
> something like a proper interface, as it were.  And a few more  
> probes (job start, job end (with return code), errors affecting job  
> scheduling, etc) perhaps might be nice too, enough
> to effectively monitor cron''s behavior.  Oh, and one other
suggested
> probe: a way to get
> cron''s idea of the timezone it is running in (although pargs could
> retrieve its TZ value,
> it would be better of all the interesting stuff could be done one  
> way).
>
> Anyone looking into adding probes to cron yet?
That would be a great RFE to submit on bugs.opensolaris.org.
Nice idea.

Adam

--
Adam Leventhal, Fishworks                        http://blogs.sun.com/ahl