[dokuwiki-teams] Re: [Infrastructure] Application update codex

  • From: Christopher Smith <chris@xxxxxxxxxxxxx>
  • To: dokuwiki-teams@xxxxxxxxxxxxx
  • Date: Sun, 30 Jan 2011 12:44:33 +0000

On 30 Jan 2011, at 08:17, Dennis Plöger wrote:

> Hi!
> 
> I think, we should talk about, what updates we install when. 
> 
> In my experience, a fast "bleeding edge" update process can lead to problems. 
> I would recommend the following:
> 
>  * High-Priority Packages (everything, that's futile to our services like 
> Apache, mysql and stuff): Wait two weeks after publication, then install
>  * Middle priority (everything system related, that's not public): three weeks
>  * Low priority (everything else) four weeks
> 
> What do you say?
> 

I don't think you mean "futile", maybe "crucial". :-)

I'm ok with your timings, except if the update corrects something that 
specifically affects us.  E.g. the recent PHP update to correct the floating 
point bug[1][2].  For updates like that one, a couple of days and a search to 
confirm no issues with the update should be sufficient.

Can we adjust the verbosity or the timing of the emails being sent by some of 
the scheduled processes?
I prefer not to receive "information" emails from the server, just "alerts" 
where some action is required.  I think its counter-productive to send emails 
at such a frequency where the default action is to ignore them.  When an 
important message comes along, it could get missed.

I'd put "apt checker", the rootkit hunter weekly db update into the information 
category.

If we are going to delay updates a couple of weeks, "application out of date" 
warnings in the rootkit hunter's daily report are information.  Just getting 
those once a week would be preferable.

And can we push forward the time that the tasks run to about 7am UTC (or at 
least the time that the emails get sent).

- Chris

[1] the php installed doesn't suffer the problem - its too old!
[2] it might be sensible to update php to 5.3.x latest, 5.3 for the improved 
garbage collection and latest as all the others have the floating point issue.

Other related posts: