[PKG-Openstack-devel] keystone: Keystone won't start, missing command

Turbo Fredriksson turbo at bayour.com
Thu Oct 13 18:04:30 UTC 2016


On Oct 13, 2016, at 4:17 PM, Thomas Goirand wrote:

> I'm not sure why it wasn't done on upgrade, really. In the postinst of
> keystone, you can read:

I can't no longer say for certain.. But this might be "my" fault..


I'm running KeepAlived with a virtual server like this:

----- s n i p -----
# => Keystone (admin)
virtual_server 10.0.4.252 35357 {
    delay_loop 10
    protocol TCP

    # Use round-robin scheduling in this example
    lb_algo rr

    # Use NAT to hide the back-end servers
    lb_kind NAT

    # Persistence of client sessions times out after 2 hours
    persistence_timeout 7200

    real_server 10.0.4.1 35357 {
        weight 1
        TCP_CHECK {
            connect_timeout 5
            connect_port 35357
        }
    }

    real_server 10.0.4.9 35357 {
        weight 1
        TCP_CHECK {
            connect_timeout 5
            connect_port 35357
        }
    }
}
----- s n i p -----

This seems to work just fine for a lot of services, but some
Openstack services don't work with this :(.

Works from local host, but not from a remote host…


So what I saw when I upgraded my second controller, it's Keystone
upgrade couldn't contact my other controller.

Testing with telnet and curl from that host, no contact. I then
checked the primary controller and couldn't see any Keystone
process running, nor could telnet/curl connect to either of
the IPs (the .1 or .252)..

Restarting Apache2, then that worked, but not from the second
controller.. So now commented out that server (to) and now
that upgrade can finish..

> BTW, I'd be happy to read about your upgrade experience.


Well… "I'm not happy". How's that for a short answer? :)


I've been using, both professionally and privately, Debian GNU/Linux
since around '94/'95 and been a Debian GNU/Linux Developer since
'97 (one of the few of the originals left :D. I can't even BEGIN
to count all the machines I've upgraded (often through two, three
releases - I've even done a three-releases upgrade, changing the
architecture in the process!) through the years and I've _NEVER_
(EVER!) had to "double check that everything still works" before!
On RedHat, CentOS etc, etc I have to do that every time, because
they ALWAYS break! But _NEVER_ Debian GNU/Linux!!

And please don't take this the wrong way, I'm very greatfull that
you're doing the job in the first place, I simply don't have the
time (or knowledge!) to do it myself.

But "you guys" needs to "step up your game" a little and _verify_
that upgrades work. And/or when you _know_ that a incompatibility
[will] happen, at least WARN the user. Preferably, do the change
_for_ the user if possible, but a simple warning would be a huge
step.

Have something "just break" without much explanation why/how is
just wrong :(.


Now, I know that this is Sid ("unstable") and things is bound to
break, it's kind'a the "rule of the game" running such bleeding
edge software (and I haven't done that in 10-15 years, and then
never a server), but if we can't even _try_ here, how are we going
to do it for "stable"?

And to try to make my previous comment sounding even less harsh,
much of the problem is me and my lack of knowledge in Openstack
and changes between different versions.

But I've never HAD to know that before! I've always just done a
"apt-get [dist-]upgrade" and everything "just works". Have _NEVER_
failed before.. Well, except when Debian GNU/Linux started linking
OpenLDAP against GnuTLS and everything just stopped working and I
had to spend almost a week to try to figure out why it didn't work
and then give up and compile it myself, because I didn't have time
to dig through all the differences and changes..


So _PLEASE_, pretty please, with a cherry on top, TRY to catch
as many of the upgrade problems as you possibly can in each and
every package.

Openstack projects are kind'a (in)famous for not caring about
"upgradability", but we're not Openstack, we're Debian GNU/Linux.
The de-facto standard on how GOOD things can and should be!
--
Att inse sin egen betydelse är som att få ett kvalster att
fatta att han bara syns i mikroskop
- Arne Anka




More information about the Openstack-devel mailing list