On Mon, May 17, 2010 at 07:55:47AM -0400, Stephen Gallagher wrote:
I don't think it makes sense to do it in be_mark_offline. I'd
prefer
that for other servers like LDAP we'd maintain affinity for the server
we currently know works.
But I might be convinced. What are everyone else's thoughts?
ok, resetting all the resolver data is not a good idea. Please find
attached a new series of patches which tries to only do what is really
needed:
0001-Revert-Create-kdcinfo-and-kpasswdinfo-file-at-startu.patch
0002-Refactor-data-provider-callbacks.patch
0003-Add-offline-callbacks.patch
0004-Refactor-krb5_finalize.patch
0005-Add-run_callbacks-flag.patch
This new patch adds a flag and a function to set it to request that
the callbacks are run the next time the resolver is called.
0006-Add-callback-to-remove-krb5-info-files-when-going-of.patch
This patch new set the run_callbacks flag only for the KDC and if
available for the kpasswd service.
0007-Krb5-locator-plugin-returns-KRB5_PLUGIN_NO_HANDLE.patch
0008-Refactor-krb5-SIGTERM-handler-installation.patch
0009-Add-krb5-SIGTERM-handler-to-ipa-auth-provider.patch
0010-Add-offline-callback-to-disconnect-global-SDAP-handl.patch
This new patch adds an offline callback to the LDAP/IPA id provider
which set the connected flag of the global SDAP handle to false.
Otherwise this connection will still be used if the transition to
the offline state was triggered by SIGUSR1 or an other provider.
bye,
Sumit