Important: upgrading a 1.23 environment

[Peter Sabaini]

On 24.07.15 05:31, Menno Smits wrote:
> The new plugin to help with stuck upgrades is done. The pull request
> is here: https://github.com/juju/plugins/pull/60. Marco: can you
> please take a look?

Heya Menno,

AFAICT the plugin will restart state servers, correct?

I'm wondering if this is safe, resp. what the potential knock-on
effects would be? For instance, will this trigger hook runs across
all units?

cheers,
peter.

> The plugin only needs to be provided the version number being
> upgraded to. It figures out the rest itself. Here's a sample of the
> plugin fixing a stuck upgrade for 3 state server environment:
> 
> 
> $ juju unstick-upgrade 1.25-alpha1.4
> 
> checking 52.12.204.107
> 52.12.204.107 is machine-1
> machine-1 fixed
> 
> checking 54.203.23.241
> 54.203.23.241 is machine-0
> machine-0 fixed
> 
> checking 54.212.170.233
> 54.212.170.233 is machine-2
> machine-2 fixed
> 
> checking 10.213.136.121
> ssh: connect to host 10.213.136.121 port 22: Connection timed out
> no machine agent found on 10.213.136.121, skipping
> 
> checking 10.255.30.118
> ssh: connect to host 10.255.30.118 port 22: Connection timed out
> no machine agent found on 10.255.30.118, skipping
> 
> checking 10.35.142.8
> ssh: connect to host 10.35.142.8 port 22: Connection timed out
> no machine agent found on 10.35.142.8, skipping
> 
> checking 52.12.204.107
> 52.12.204.107 is machine-1
> machine-1 has already been fixed, skipping
> 
> fix complete
> 
> 
> - Menno
> 
> 
> On 24 July 2015 at 10:51, Menno Smits <menno.smits at canonical.com
> <mailto:menno.smits at canonical.com>> wrote:
> 
> 
>         I am interested in the 1.24 aspect. Can you elaborate on:
> 
>         1. What exact 1.24 release is affected?
> 
> 
>     The issue was fixed in 1.24.3 so 1.24.0 and 1.24.1 are affected.
> 
>     Here's the ticket: https://bugs.launchpad.net/juju-core/+bug/1468653
> 
>         2. To what degree is it affected? You said "may" also get stuck.
> 
> 
>     It depends on whether certain leadership related API requests
>     had been made and whether the lease worker had restarted while
>     those requests were active. The more units in the environment,
>     the more likely the problem is to occur.
> 
> 
> 
> 
> 
> 
> 
> 
> 


-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 801 bytes
Desc: OpenPGP digital signature
URL: <https://lists.ubuntu.com/archives/juju/attachments/20150724/d1dc7e75/attachment.pgp>