https://bugzilla.wikimedia.org/show_bug.cgi?id=63382
Alexandros Kosiaris <akosia...@wikimedia.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Status|PATCH_TO_REVIEW |ASSIGNED
--- Comment #16 from Alexandros Kosiaris <akosia...@wikimedia.org> ---
I would argue the confusion stems from the "real" and "fake" attributions.
There is no "fake" nor "real" labsdb1004. There is a master postgresql server
(labsdb1004) and a hot standby (labsdb1005). That means that all applications
should use labsdb1004.
labsdb1005 is there so that the operations team can recover the service faster
in case of labsdb1004 hardware failure. In which case, apart from the
(hopefully minimal) downtime, during which we will be switching over, users and
applications should not notice anything and should not change any setting.
The fact that labsdb1005 is capable of answering read-only queries is a
conscious design choice, but if it is really causing all this confusion, that
can be undone. In fact, just ignoring/forgetting about the existence of
labsdb1005 might actually be better for users.
That being said, a separate DNS will indeed provide the ops team with greater
flexibility when the unfortunate time of a labsdb1004 hardware failure comes.
It will however provide no extra benefit to users. I will look into it to see
how we can do it and still be consistent with the rest of Labs.
--
You are receiving this mail because:
You are on the CC list for the bug.
_______________________________________________
Wikibugs-l mailing list
Wikibugs-l@lists.wikimedia.org
https://lists.wikimedia.org/mailman/listinfo/wikibugs-l
