[Date Prev][Date Next]
RE: multiple slapd Accessing one data area
If you're using back-ldbm I'm pretty sure that none of the available
database libraries (ndbm, gdbm, bdb) support multiple access across NFS.
None of (ndbm, gdbm) are even suited for multiple access on a single host.
gdbm runs in N-readers/1-writer mode; since back-ldbm opens the database in
RW mode there's no way multiple slapd instances could access the database.
BDB supports locking on a per-page basis (and so, perhaps it will allow
multiple slapds on a single database) but in general its lock system lives
in shared memory, which also does not propagate across NFS. (The BDB
documentation explicitly says not to store BDB databases in NFS, for this
and other reasons.)
So, if you have a working NFS implementation (with working lockd) then gdbm
shouldn't allow you to even open the database multiple times. If you don't
have a working lockd, then you will probably corrupt the database very
quickly. For bdb, you will just corrupt the database.
Note that bdb (at least in version 3, I dunno about earlier versions) has an
RPC interface that could allow you to run multihost access without using
NFS. But there's nothing in slapd that will set this up for you.
-- Howard Chu
Chief Architect, Symas Corp. Director, Highland Sun
Symas: Premier OpenSource Development and Support
> -----Original Message-----
> From: owner-openldap-software@OpenLDAP.org
> [mailto:owner-openldap-software@OpenLDAP.org]On Behalf Of rj
> Sent: Friday, November 30, 2001 9:41 PM
> To: openldap-software@OpenLDAP.org
> Subject: multiple slapd Accessing one data area
> Hi Dear List,
> is it possible in openldap1.2.9 for multiple slapd on different
> systems to access one data area ( all the .dbb files) if yes
> where do i specify it.
> one option that i can think of now is mounting via nfs,
> and what i have done is
> on machine A i have ldap server with data in /aaa directory. now
> from system B i mount the /aaa of machine A in Read/Write Mode
> and in the slapd.conf file i give the directory as the mount
> point. when i start the servers on both the machines i am able to
> perform search successfully, but any modification, add or del are
> only getting reflected when i restart the server. is it the
> problem with the File system r the underlying database "locking"
> mechanism. can someone make this clear to me.
> Or do we have other options.
> Thanks in advance