<div dir="ltr"><br><div class="gmail_extra"><br><br><div class="gmail_quote">On Thu, Apr 4, 2013 at 10:57 AM, William Reade <span dir="ltr"><<a href="mailto:william.reade@canonical.com" target="_blank">william.reade@canonical.com</a>></span> wrote:<br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div class="im">On Thu, 2013-04-04 at 15:40 +0100, Tom Haddon wrote:<br>
> The particular case here is that we don't actually have a relation<br>
> error, we just want to change the implementation of the hook. Can you<br>
> run "juju resolved --retry unit_name [relation_name]" even if there's no<br>
> error?<br>
<br>
</div>Ah, that's a bit trickier. but... with judicious use of the various<br>
relation-* hook tools, you should be able to retrieve all the<br>
information that the past hooks had access to, and should thus be able<br>
to reconstruct anything you had previously constructed from that info.<br>
<br>
FWIW, I should point out that juju-core's resolved has no<br>
[relation_name], because relation errors have been promoted to unit<br>
errors: as soon as any hook fails we stop and wait, rather than running<br>
further hooks in a potentially-broken context.<br></blockquote><div><br></div><div><br></div><div style>Just to be clear on behavior, while pyjuju does mark relation hook errors separately, it also stops all hook execution on the unit on a relation hook error.</div>
<div style><br></div><div style>cheers,</div><div style><br></div><div style>Kapil</div><div><br></div></div><br></div></div>