[7560] in Release_7.7_team
Re: Permission to use update hook?
daemon@ATHENA.MIT.EDU (Jonathan Reed)
Wed Jul 27 08:06:37 2011
Mime-Version: 1.0 (Apple Message framework v1084)
Content-Type: text/plain; charset=us-ascii
From: Jonathan Reed <jdreed@MIT.EDU>
In-Reply-To: <alpine.DEB.2.00.1107262137350.4814@tyger.mit.edu>
Date: Wed, 27 Jul 2011 08:06:29 -0400
Cc: Jonathon Weiss <jweiss@mit.edu>, release-team@mit.edu
Message-Id: <E55A57F1-8913-43F3-B0ED-6ED1710CBD07@mit.edu>
To: Geoffrey Thomas <geofft@mit.edu>
Content-Transfer-Encoding: 8bit
Hrm. I'm on the fence about this. On one hand, I think I agree with Geoff, in that I'd like to see what happens if we effectively triple the number of machines talking simultaneously (although desync'd, each auto-update would now be talking to the APT repo 2 more times than it normally would -- once for the aptitude update, once for the aptitude install). In fact, I'd argue that the auto-updater should always attempt to auto-update itself first, so that we don't have to invoke the recovery hook for things like this.
OTOH, I'm not sure that hammering scripts.mit.edu is a useful test, and I agree that the update hook is disaster recovery, and therefore we should code it so that it recovers from the specific disaster we're trying to deal with.
Given that (a) it's summer, (b) the worst that happens is that the update hook fails and Nagios is sad until the next update, (c) we can pull it if it starts exploding, I think I'll go with the following at 1:30pm today if nobody has any other concerns. I will pull it at 4:00pm today (after the last natty -dev machine has taken its update).
{{{
#!/bin/sh
aptitude update
aptitude -y install debathena-auto-update
}}}
Should we use apt-get?
-Jon
On Jul 26, 2011, at 9:39 PM, Geoffrey Thomas wrote:
> I kind of actually want to see this run on all the machines so that we know now (not during the semester and otherwise at a generally chill time) if running a command of this form on the clusters all at once is a bad idea.
>
> --
> Geoffrey Thomas
> geofft@mit.edu
>
> On Tue, 26 Jul 2011, Jonathon Weiss wrote:
>
>>
>> Is it worth coding this to explicitly only run on natty machines? I
>> realize that the suggested fix should be a no-op everywhere else, but
>> the paranoid in me says that scripts like this should touch as little as
>> possible.
>>
>> Jonathon
>>
>>
>> Jonathan Reed <jdreed@MIT.EDU> wrote:
>>
>>> Due to #987, there are now a number (3) of -development machines which can't take an update. This would be an ideal time to test out our update hook in real life, rather than waiting for a last resort. I propose trying out the following update hook:
>>>
>>> """
>>> #!/bin/sh
>>>
>>> aptitude install debathena-auto-update
>>> """
>>>
>>> This should be a no-op on production machines, and should pull in the new debathena-auto-update on -dev machines.
>>>
>>> Thoughts? We can also decide that the update hook should really truly only be reserved for the end of the world.
>>>
>>> -Jon
>>
