[Labs-l] Partial (but dramatic) labs outage on Tuesday: 2015-02-24 1500UTC-1800UTC

[Petr Bena]

RANT 2

Why don't we investigate what is taking so much space there? AFAIK
it's 30TB storage, it shouldn't be filling up rapidly, isn't that just
some broken tool that infinitely writes garbage to /data/project?

On Sat, Feb 21, 2015 at 11:27 AM, Petr Bena <benapetr at gmail.com> wrote:
> RANT
>
> If puppet was written in proper language (C++) we wouldn't need more RAM :P
>
> On Fri, Feb 20, 2015 at 5:07 PM, Ricordisamoa
> <ricordisamoa at openmailbox.org> wrote:
>> Thank you.
>> I (and probably many others) would like someone from the Ops team to
>> elaborate on the uptime and general reliability Labs (especially Tools) is
>> supposed to have, and for what kind of services it is suitable for, to
>> prevent future misunderstandings in regards to loss of important work, etc.
>>
>> Il 19/02/2015 22:00, Andrew Bogott ha scritto:
>>
>>> It is with a heavy heart that I must share the news of an upcoming Labs
>>> maintenance window.
>>>
>>> The labs NFS store (which you probably know as /data/project) is filling
>>> up rapidly and we need to add more drives.  By weird coincidence the actual
>>> physical space for that server in the datacenter is ALSO filling up, so
>>> Chris Johnson has graciously agreed to spend his day re-shuffling servers in
>>> order to make space for the new diskshelf.  This involves lots of unplugging
>>> and replugging and amounts to the fact that the NFS server will need to be
>>> turned off for several hours.
>>>
>>> During this window Chris will take care of another long-deferred
>>> maintenance task -- he's putting more RAM into the labs puppet master,
>>> virt1000.
>>>
>>> What will break:
>>>
>>> - Shared storage for all labs and tools instances.  That includes volumes
>>> like /data/project, /public/dumps, /data/scratch, /home
>>>
>>> - Logins to all instances running ubuntu Precise.  (Trusty hosts will
>>> /probably/ still support logins.)
>>>
>>> - Login to wikitech and manipulation of instances.
>>>
>>> What won't break:
>>>
>>> - Labs instances will continue to run
>>>
>>> - Tasks running on instances will continue to run; those that don't rely
>>> on shared storage should be fine.
>>>
>>> - Web proxies should keep working, if the services they support aren't
>>> relying on shared storage.
>>>
>>> What will get better:
>>>
>>> - More storage space!
>>>
>>> - Fewer problems with dumps filling up NFS (which is basically the same as
>>> 'more storage space'.
>>>
>>> - More reliable puppet runs and fewer outages with miscellaneous OpenStack
>>> services (which also run on virt1000)
>>>
>>> I apologize in advance for this downtime.  Don't hesitate to contact me or
>>> Coren either here or on IRC with advice about how to harden your tool
>>> against this upcoming outage.  We will also be available on IRC during and
>>> after the outage to help revive things that are angry about the timeouts.
>>>
>>> -Andrew
>>>
>>> _______________________________________________
>>> Labs-l mailing list
>>> Labs-l at lists.wikimedia.org
>>> https://lists.wikimedia.org/mailman/listinfo/labs-l
>>
>>
>>
>> _______________________________________________
>> Labs-l mailing list
>> Labs-l at lists.wikimedia.org
>> https://lists.wikimedia.org/mailman/listinfo/labs-l