On Wed, Sep 8, 2021 at 2:36 AM Erdősi Péter <erdosi.peter(a)kifu.gov.hu>
wrote:
Hello!
We're started the upgrade one of our oVirt clusters to the recent 4.4
minor version (4.4.8.5-1.el8) and we found a bug in the recent change of
the aaa-ldap plugin.
This bug came along after the IPv4/IPv6 selection introduced in
ovirt-engine-extension-aaa-ldap 1.4.4
We've dig down in the rabbit hole, and it looks like, our DNS solution is
just not compatible with the plugin after this commit:
https://github.com/oVirt/ovirt-engine-extension-aaa-ldap/commit/4c0f2e72d...
I don't know, how the industry actually use CNAME records, but we have a
rule, that we use independet DNS names for services, and machines itself.
This become handy, when you migrate/upgrade services, like LDAP to another
machine, and I think, this should be supported.
Our LDAP server service domain is actually ldap-master.niif.hu and
ldap.niif.hu, but in real, there are two servers, which actually have
their own FQDNs.
The actual DNS structure:
host -t CNAME ldap-master.niif.hu
ldap-master.niif.hu is an alias for elm.niif.hu.
host -t CNAME ldap.niif.hu
ldap.niif.hu is an alias for holly.ldap.einfra.hu.
host elm.niif.hu
elm.niif.hu has address 193.225.14.175
elm.niif.hu has IPv6 address 2001:738:0:701::3
host holly.ldap.einfra.hu
holly.ldap.einfra.hu has address 193.224.92.6
holly.ldap.einfra.hu has IPv6 address 2001:738:0:431::6
The commit above (as far as I understand) only tries to resolve A and AAAA
records in DNS, and drop the connection if it not found. Of course, the
certificate only have ldap-master and ldap.niif.hu in it, so using holly
end elm does not solve the problem (also, if the service will be migrated,
the service domain will be kept, but not the machine's FQDN, since we
cannot afford to shut down one of our LDAP server for a migration windows.
We've tried to downgrade the package to 1.4.3, which is works fine.
The actual error looks like this (engine.log)
2021-09-07 15:53:09,833+02 WARN
[org.ovirt.engine.extension.aaa.ldap.AuthnExtension] (default task-1) []
[ovirt-engine-extension-aaa-ldap.authn::NIIFLdap-authn] Cannot initialize
LDAP framework, deferring initialization. Error: An error occurred while
attempting to connect to server ldap-master.niif.hu:636:
IOException(LDAPException(resultCode=91 (connect error), errorMessage='An
error occurred while attempting to establish a connection to server
ldap-master.niif.hu/193.225.14.175:636: UnknownHostException(
ldap-master.niif.hu), ldapSDKVersion=4.0.14,
revision=c0fb784eebf9d36a67c736d0428fb3577f2e25bb'))
2021-09-07 15:53:09,833+02 ERROR
[org.ovirt.engine.core.sso.servlets.InteractiveAuthServlet] (default
task-1) [] Internal Server Error: An error occurred while attempting to
connect to server ldap-master.niif.hu:636:
IOException(LDAPException(resultCode=91 (connect error), errorMessage='An
error occurred while attempting to establish a connection to server
ldap-master.niif.hu/193.225.14.175:636: UnknownHostException(
ldap-master.niif.hu), ldapSDKVersion=4.0.14,
revision=c0fb784eebf9d36a67c736d0428fb3577f2e25bb'))
2021-09-07 15:53:09,833+02 ERROR
[org.ovirt.engine.core.sso.service.SsoService] (default task-1) [] An error
occurred while attempting to connect to server ldap-master.niif.hu:636:
IOException(LDAPException(resultCode=91 (connect error), errorMessage='An
error occurred while attempting to establish a connection to server
ldap-master.niif.hu/193.225.14.175:636: UnknownHostException(
ldap-master.niif.hu), ldapSDKVersion=4.0.14,
revision=c0fb784eebf9d36a67c736d0428fb3577f2e25bb'))
2021-09-07 15:53:09,854+02 ERROR
[org.ovirt.engine.core.aaa.servlet.SsoPostLoginServlet] (default task-1) []
server_error: An error occurred while attempting to connect to server
ldap-master.niif.hu:636: IOException(LDAPException(resultCode=91
(connect error), errorMessage='An error occurred while attempting to
establish a connection to server ldap-master.niif.hu/193.225.14.175:636:
UnknownHostException(ldap-master.niif.hu), ldapSDKVersion=4.0.14,
revision=c0fb784eebf9d36a67c736d0428fb3577f2e25bb'))
As we tried to run the setup tool, that is also looks broken (after that,
we've copied the required files to /etc/ovirt-engine/aaa and
/etc/ovirt-engine/extensions.d/ from other, working hosted engine) so we've
tested the plugin itself, and the setup too, but no luck.
I think, this (CNAME in DNS) should be working with the plugin.
Could you please investigate this issue? (we're here to help test the
repaired version/patch, if needed, but not feel the knowledge to create the
patch ourself)
From very casual skimming of the patch, it seems to me like you should be
able to restore the previous behavior by setting in your conf file
'pool.default.socketfactory.resolver.detectIPVersion = false'. Can you
please try that?
Adding Martin.
Best regards,
--
Didi