[PATCH] LDAP: Try next failover server on any error

[Stephen Gallagher]

Fixes https://fedorahosted.org/sssd/ticket/1101

Previously, we were only trying the next failover server if we got back
ETIMEDOUT (which was a relic of using ldap_initialize()). Now that we're
directly connecting to the socket manually, we need to ensure that we
handle all possible socket errors.

With this fix, we will now just treat ALL errors as requiring us to try
the next server.

[Jakub Hrozek]

On Mon, Nov 28, 2011 at 02:01:44PM -0500, Stephen Gallagher wrote:
> Fixes https://fedorahosted.org/sssd/ticket/1101
>
> Previously, we were only trying the next failover server if we got back
> ETIMEDOUT (which was a relic of using ldap_initialize()). Now that we're
> directly connecting to the socket manually, we need to ensure that we
> handle all possible socket errors.
>
> With this fix, we will now just treat ALL errors as requiring us to try
> the next server.

Ack

[Jan Zelený]

> On Mon, Nov 28, 2011 at 02:01:44PM -0500, Stephen Gallagher wrote:
> > Fixes https://fedorahosted.org/sssd/ticket/1101
> >
> > Previously, we were only trying the next failover server if we got back
> > ETIMEDOUT (which was a relic of using ldap_initialize()). Now that we're
> > directly connecting to the socket manually, we need to ensure that we
> > handle all possible socket errors.
> >
> > With this fix, we will now just treat ALL errors as requiring us to try
> > the next server.
>
> Ack

Pushed to master, I'm not sure whether to push this to sssd-1-5 and sssd-1-6 
as well.

Jan

[Stephen Gallagher]

On Tue, 2011-11-29 at 09:45 +0100, Jan Zelený wrote:
> > On Mon, Nov 28, 2011 at 02:01:44PM -0500, Stephen Gallagher wrote:
> > > Fixes https://fedorahosted.org/sssd/ticket/1101
> > >
> > > Previously, we were only trying the next failover server if we got back
> > > ETIMEDOUT (which was a relic of using ldap_initialize()). Now that we're
> > > directly connecting to the socket manually, we need to ensure that we
> > > handle all possible socket errors.
> > >
> > > With this fix, we will now just treat ALL errors as requiring us to try
> > > the next server.
> >
> > Ack
>
> Pushed to master, I'm not sure whether to push this to sssd-1-5 and sssd-1-6 
> as well.

Pushed to sssd-1-6 and sssd-1-5 as well. This was a regression from the
behaviour before we switched to maintaining the socket fd ourselves.

[Jan Zelený]

> On Tue, 2011-11-29 at 09:45 +0100, Jan Zelený wrote:
> > > On Mon, Nov 28, 2011 at 02:01:44PM -0500, Stephen Gallagher wrote:
> > > > Fixes https://fedorahosted.org/sssd/ticket/1101
> > > >
> > > > Previously, we were only trying the next failover server if we got
> > > > back ETIMEDOUT (which was a relic of using ldap_initialize()). Now
> > > > that we're directly connecting to the socket manually, we need to
> > > > ensure that we handle all possible socket errors.
> > > >
> > > > With this fix, we will now just treat ALL errors as requiring us to
> > > > try the next server.
> > >
> > > Ack
> >
> > Pushed to master, I'm not sure whether to push this to sssd-1-5 and
> > sssd-1-6 as well.
>
> Pushed to sssd-1-6 and sssd-1-5 as well. This was a regression from the
> behaviour before we switched to maintaining the socket fd ourselves.

I thought so, I just didn't want to break something in stable branches.

Thanks
Jan