Indeed, routine Phabricator maintenance is scheduled to take place every Wednesday at 0:00 UTC. This maintenance window is not always utilized as we only make changes once or twice each month. When there is a maintenance task, typical downtime is less than 5 minutes. When nothing needs to be done for a given week, I try to at least !log the absence of any maintenance. Realistically though I often forget to do that.
This week the planned work involves a lot of changes so expect a few minutes of downtime while we switch everything over. Then we will take a few more minutes to validate that the new system is working as expected, if not we will roll back to the old one.
Optimistically I'd guess that the whole process takes 10 to 15 minutes with downtime closer to 5 minutes.
On Wed, May 22, 2019 at 2:02 AM Lewis Cawte via Wikitech-l < wikitech-l@lists.wikimedia.org> wrote:
To answer the first part of your question, Pine, Daniel's email was sent at 01:52 UTC on the 22nd. So tomorrow would refer to 00:00 on the 23rd.
-- Lewis Cawte (Lcawte)
On Wed, 22 May 2019 at 07:26, Pine W wiki.pine@gmail.com wrote:
Hi, thanks for the notice. Can you specify whether "tomorrow" refers to
22
May UTC or 23 May UTC, and how long the maintenance window is likely to
be?
Some of us may want to read or write a Phabricator <strike>complaint</strike> task during the maintenance window, and if we know when to expect Phabricator may be temporarily unavailable then we
can
plan accordingly.
Pine
( https://meta.wikimedia.org/wiki/User:Pine )
On Tue, May 21, 2019, 17:52 Daniel Zahn dzahn@wikimedia.org wrote:
Tomorrow we have planned a major upgrade of the Phabricator prod
server.
Starting from around 00:00 UTC (after Evening SWAT) we want to switch
from
phab1001 to phab1003 as the production server.
This will involve:
- switching from jessie to stretch
- switching from mod_php to php-fpm
- switching from PHP 5 to PHP 7.2
- upgrading apache from 2.4.10 to 2.4.25
We are also hoping it will finally fix the httpd memory leaks we have
been
seeing on the current server for a long time and which require periodic service restarts.
In case we run into any unexpected issue we will simply fall back to phab1001 and keep running as before.
https://phabricator.wikimedia.org/T221389 https://phabricator.wikimedia.org/T182832 https://phabricator.wikimedia.org/T190568 https://phabricator.wikimedia.org/T151070
Cheers,
Daniel and Mukunda
-- Daniel Zahn dzahn@wikimedia.org Operations Engineer _______________________________________________ Wikitech-l mailing list Wikitech-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/wikitech-l
Wikitech-l mailing list Wikitech-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/wikitech-l
Wikitech-l mailing list Wikitech-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/wikitech-l