[tor-bugs] #33477 [Internal Services/Tor Sysadmin Team]: automate host retirement / decommissionning procedure

[Tor Bug Tracker & Wiki]

#33477: automate host retirement / decommissionning procedure
-------------------------------------------------+-------------------------
 Reporter:  anarcat                              |          Owner:  anarcat
     Type:  defect                               |         Status:
                                                 |  assigned
 Priority:  Medium                               |      Milestone:
Component:  Internal Services/Tor Sysadmin Team  |        Version:
 Severity:  Normal                               |     Resolution:
 Keywords:                                       |  Actual Points:
Parent ID:                                       |         Points:
 Reviewer:                                       |        Sponsor:
-------------------------------------------------+-------------------------
Description changed by anarcat:

Old description:

> [we're removing a surprising number of servers. in the last few months,
> we're retired about half a dozen servers. that procedure is currently
> entirely manual, and quite error-prone, especially because of at(1) jobs
> errors or omitted steps. backup removal can be forgotten or typo'd and
> we've forgotten to remove entries in the spreadsheet or nagios a few
> times.
>
> we should automate this process.
>
> right now the documentation is in the
> [https://help.torproject.org/tsa/howto/retire-a-host/ retire-a-host]
> page, but i somehow got into the habit of calling tickets `decomission
> host FOO`. so we need to decide between "retire" and "decommission" as a
> naming convention, before we start writing code.
>
> the two contestants are:
>
>  * retire, retirement, retiring, retired
>  * decommission, decommissioning procedure, decommissioning,
> decommisssioned
>
> a quick IRC survey indicates people favor the former family because it's
> shorter and more familiar.
>
> decommission was seen as a "nice word". it's also less ambiguous:
> "retire" can also refer to a user, and it could also imply the host
> sticks around for a while and rant about pain in his lower backs and
> babble nonsense when guests are around, just to embarrass us. the problem
> with decommission is that I can't spell it to save my life. it also
> doesn't have a "name" (an action, like "retirement") so it makes it
> sometimes awkward to refer to.
>
> so we favor converging over "retire" for now.

New description:

 [we're removing a surprising number of servers. in the last few months,
 we're retired about half a dozen servers. that procedure is currently
 entirely manual, and quite error-prone, especially because of at(1) jobs
 errors or omitted steps. backup removal can be forgotten or typo'd and
 we've forgotten to remove entries in the spreadsheet or nagios a few
 times.

 we should automate this process. this also has the added benefit of
 simplifying the ''migration'' process: we have a ton of servers that need
 to move from libvirt into the ganeti cluster, and part of that process
 involves retiring the old copy of the server.

 right now the documentation is in the
 [https://help.torproject.org/tsa/howto/retire-a-host/ retire-a-host] page,
 but i somehow got into the habit of calling tickets `decomission host
 FOO`. so we need to decide between "retire" and "decommission" as a naming
 convention, before we start writing code.

 the two contestants are:

  * retire, retirement, retiring, retired
  * decommission, decommissioning procedure, decommissioning,
 decommisssioned

 a quick IRC survey indicates people favor the former family because it's
 shorter and more familiar.

 decommission was seen as a "nice word". it's also less ambiguous: "retire"
 can also refer to a user, and it could also imply the host sticks around
 for a while and rant about pain in his lower backs and babble nonsense
 when guests are around, just to embarrass us. the problem with
 decommission is that I can't spell it to save my life. it also doesn't
 have a "name" (an action, like "retirement") so it makes it sometimes
 awkward to refer to.

 so we favor converging over "retire" for now.

--

--
Ticket URL: <https://trac.torproject.org/projects/tor/ticket/33477#comment:2>
Tor Bug Tracker & Wiki <https://trac.torproject.org/>
The Tor Project: anonymity online