[Date Prev][Date Next] [Chronological] [Thread] [Top]

Re: back-bdb cache and back-monitor

> Is there any thought to developing (or is there already and I've missed
> it) a way to see if the cachesize (as specified in slapd.conf) is
> sufficient for the workload?  What I'm thinking of is a query to the
> back-monitor which would return the cache hit/miss numbers.  Is that
> possible?  Are they even calculated?

Currently, this is not possible.  I haven't looked at it in detail, but I
think there should not be any relevant technical problem.  However, let me
repeat that I haven't looked ad back-bdb's cache close enough to state
this for sure, nor I'm committing to looking at it :).

Currently, there's some work on back-monitor that should make it easy, by
allowing external code to register and update entries in back-monitor. 
back-bdb could exploit this by registering a "cn=Cache" entry under each
database instance of the bdb backend and registering a callback that, upon
monitor entry read request, updates the desired figure by referencing the
data stored in back-bdb itself.  Then, it's just a matter of recording the
desired figures, with appropriate concurrency protection.


Pierangelo Masarati

    SysNet - via Dossi,8 27100 Pavia Tel: +390382573859 Fax: +390382476497