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

Re: cn=config and mdb questions



On 08/12/2014 19:03, Dieter Klünter wrote:
Am Mon, 08 Dec 2014 15:05:28 +1000
schrieb Da Rock <openldap-technical@herveybayaustralia.com.au>:

Aside from other issues (which haven't really been solved... just
possibly worked around :/), I've been doing some tinkering and have
raised some questions which I'm hoping will have answers to.

1. cn=config and friends: Looking at the schema and methods used to
implement this config type, I got to considering that I could (in
theory) name my databases such as olcDatabase=MyLDAPDatabase (or the
like), rather than bdb/hdb/mdb. I then found that I couldn't and then
looked it up to see why - I found that it had a naming restriction,
but not why. Aside from possible limits within the programming as it
stands now, is there some semantical issue why this is the case?
Wouldn't possibly be easier to determine which database is which if
the administrator was able to name them as they wish?
OpenLDAP provides a multitude of backends/databases
http://www.openldap.org/faq/data/cache/1165.html
you have to define the database required somewhere, and that is per
definition the configuration parameter database or oclDatabase.
Yes guys, this I understand and had read the manual. But within the definition in olc format is the objectclass that defines the type - ie olcBdbConfig. Therefore, I considered the fact that, seeing as the system has an easy means to know which db type is used based on schema, then an admin could theoretically name the olcDatabase attr to a more memorable, convenient name. So I wondered if anything other than programming might be holding it back.

Nothing fire people up, just my simple curiosity :)


2. mdb: I know this is the way we're expected to go with ldap now,
but what is the point of a db that can't grow as needed? It seems
rather limiting (and old school - remember early unix programming on
the old IBM mainframes and DOS for the PC's later with the 640k
limit), so why the push in this direction? Or is it expected that
this will be resolved in the future?
I think there is some misunderstanding, slapd-mdb(5) is quite clear:

maxsize <bytes>
Specify  the maximum size of the database in bytes. A
memory map of this size is allocated at startup time and the database
will not be allowed to grow beyond this size.
You are still free to resize the database to yout requirements.
I thought that was the case, but couldn't find it again. There also seems to be a question mark even within the docs I read on it as well, so I thought it would best to make sure too. In fact even my man pages have a note warning about resizing limitations given heavy load.

Another thought to that is that it seems rather cumbersome to have to manually resize; could very well catch you unawares.