On Thu, May 22, 2014 at 01:55:44PM -0700, Daniel Jung wrote:
I was curious on how the servers with same priority with weights were
implemented, the wording in RFC on this algorithm was a bit hard to
visualize for me and whether this was strictly followed.
You can see the implementation of the weight selection here:
Even if you're not a C programmer, maybe the comments will show how we
follow the RFC. The intent of the code is to share the load from several
clients according to the sum of the weights on the same priority level.
So if you had two servers A and B with the same priority with weights
of 70 and 30 respectively, 70% of clients should select server A and 30%
should select server B.
Also, at which
timeout setting is applied for cases where selected server is not reachable
and next server is selected and connected?
Would this be same timeout
setting when using multiple servers with URI instead of DN?
There are several timeouts at play, depending on how exactly the server
is unreachable and what the provider is. For DNS resolution itself,
dns_resolver_timeout is applied. Once you have an IP address and start
connecting to an LDAP server, we try for ldap_network_timeout seconds.
There are different timeouts for Kerberos, you can see them all in the
sssd-ldap and sssd-krb5 man pages.
Thanks for all the information.
On Wed, May 21, 2014 at 3:14 AM, Jakub Hrozek <jhrozek(a)redhat.com> wrote:
> On Tue, May 20, 2014 at 11:01:42PM -0700, Daniel Jung wrote:
> > thanks for the response guys. just one more question on the topic of SRV
> > records, does sssd implementation folllow srv rfc closely? would i need
> > dig into the code to find this?
> As far as I know it does, the code was modeled after the RFC. Is there
> any particular functionality that you are concerned about?
> sssd-users mailing list
sssd-users mailing list