kdb5_util fails to load propagated database under heavy load

[Christopher Odenbach]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1


Hi,

a few days ago we upgraded our KDC server from Debian squeeze to Debian
jessie. Since then the first slave KDC sometimes reported an error
during database propagation, maybe two or three times a day, probably
during times of heavy usage.

To get more information I started kpropd in debug mode. When the error
occurs I get the following lines:

[...]
Full propagation transfer started.
Full propagation transfer finished.
calling kdb5_util to load database
Load PID is 20249
kdb5_util: Resource temporarily unavailable while making newly loaded
database live
kdb5_util: Invalid argument while deleting bad database
/var/lib/krb5kdc/principal
kpropd: /usr/sbin/kdb5_util returned a bad exit status (1)
Database load process for full propagation completed.

- From the source I found out that this error is thrown at

src/kadmin/dbutil/dump.c:

        ret = krb5_db_promote(util_context, db5util_db_args);
        /* Ignore a not supported error since there is nothing to do
about it
         * anyway. */
        if (ret != 0 && ret != KRB5_PLUGIN_OP_NOTSUPP) {
            com_err(progname, ret,
                    _("while making newly loaded database live"));
            goto error;
        }


So from my understanding there must be some kind of locking problem
between krb5_db_promote() and the normal KDC operations.

Has anyone experienced these error messages as well?

Thanks,

Christopher

- -- 
======================================================
    Dipl.-Ing. Christopher Odenbach
    Zentrum fuer Informations- und Medientechnologien
    Universitaet Paderborn
    Raum N5.311
    odenbach at uni-paderborn.de
    Tel.: +49 5251 60 5315
======================================================

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2

iD8DBQFWzIVuhxiCJKeLY0IRAit7AJ0T/U5Upb1+3lTDA62iqEtLS8afggCfTQNt
M5oLBw77Bm5yAktAcdBtRvc=
=ymXq
-----END PGP SIGNATURE-----