Important maintenance bug

classic Classic list List threaded Threaded
2 messages Options
Reply | Threaded
Open this post in threaded view
|

Important maintenance bug

Leslie Carr
https://bugzilla.wikimedia.org/show_bug.cgi?id=46428

If one of the php-knowledgable peeps can take a look at this (sadly my
php-foo is quite weak).

--
Leslie Carr
Wikimedia Foundation
AS 14907, 43821
http://as14907.peeringdb.com/

_______________________________________________
Wikitech-l mailing list
[hidden email]
https://lists.wikimedia.org/mailman/listinfo/wikitech-l
Reply | Threaded
Open this post in threaded view
|

Re: Important maintenance bug

Brion Vibber
On Thu, Mar 21, 2013 at 11:46 AM, Leslie Carr <[hidden email]> wrote:
> https://bugzilla.wikimedia.org/show_bug.cgi?id=46428
>
> If one of the php-knowledgable peeps can take a look at this (sadly my
> php-foo is quite weak).

Hmm... ok this explains why I couldn't find Wikipedia Zero-related
memcache entries when testing from fenari last night. :(

It sounds like maintenance scripts could have a serious split-brain
problem if they're being run in pmtpa while the web servers are in
eqiad, if each cluster uses its own local memcached servers.

Either updates need to be proxied to both clusters (race conditions?)
or it should only be possible to run maintenance scripts in the same
cluster that's running the web servers. The former sounds scary and
latency-intense, while the latter requires more configuration on the
eqiad bastion hosts so MW scripts can be run. :(

-- brion

_______________________________________________
Wikitech-l mailing list
[hidden email]
https://lists.wikimedia.org/mailman/listinfo/wikitech-l