Previously we had an issue that every other update_active_status
request for a particular realm would result in doing the expensive
query to compute the list of active users in that realm. It turned
out this was because on every update_active_status request, we'd queue
an event that would have the effect of clearing the cache, even if
nobody's tatus changed. This fixes that issue, by only clearing the
cache for a realm if someone's status actually changed (or the 60s
timeout expires).
(imported from commit d5b829fe255a31c8cecb58458738f1e72a2cf6de)
The refactoring to use the cache_get() method incorrectly didn't
remove the addition of KEY_PREFIX inside cache_with_key. The result
was that the KEY_PREFIX was being added twice, once by cache_with_key
and again inside cache_get.
This had the impact of causing pointer saves to not take effect,
because our attempts to update the memcached cache when we save the
UserProfile object were using the correct cache key, but the actual
code reading values out of the caceh wasn't.
(imported from commit dcea000833f00622bdc0249488de3b186a7417b2)
Otherwise code paths that use those keys, like get_old_messages, will
incorrectly use the prefix-included keys.
This bug in our KEY_PREFIX system results in our memcached caching for
get_old_messages always missing.
(imported from commit 506c13e06d6f266596ead0b381c324c256e576c3)
Some cache keys used by Django (like sessions) will not have the key
prefixes, but those values shouldn't change across most restarts.
(imported from commit 2fe61028111fe9d5700432214a611b3341412654)
Since we log to statsd our cache time lookups by cache key, using a unique
tweet id for each lookup was just filling up our cache without being useful.
Also, log database cache lookups in a further namespace to distinguish between
memcached caches
(imported from commit a2a16b777fb7ab8cd066feee7344f9c8a3c107f5)
After this change, the memcached time consumed by doing
get_old_messages for 200 and 1000 messages respectively now look like
this:
200 63ms (mem: 6ms/3) (db: 4ms/2q) /json/get_old_messages
200 178ms (mem: 67ms/2) (db: 6ms/1q) /json/get_old_messages
which might help explain where the time is going on prod for some of
our slower queries.
(imported from commit b8fe83b175914b6796922a65a1c5537f4e7a9429)
On my laptop, this saves about 80 milliseconds per 1000 messages
requested via get_old_messages queries. Since we only have one
memcached process and it does not run with special priority, this
might have significant impact on load during server restarts.
(imported from commit 06ad13f32f4a6d87a0664c96297ef9843f410ac5)
See PEP 328[1] for details. This feature was introduced in Python 2.5 and
will become mandatory in Python 3.
[1]: http://www.python.org/dev/peps/pep-0328
(imported from commit 7444eeba8a08d5f91b94c7921848f2274979bd76)
I think all that one needs to do to deploy this commit is on developer
laptops, run `generate-fixtures --force`.
(imported from commit 34916341435fef0875b5a2c7f53c2f5606cd16cd)
This saves 2 database queries per user in the huddle when sending the
first message to a particular huddle.
(imported from commit f71aa32df846fb4b82651a93ff9608087ffcaa5a)
This fixes an experienced bug where you couldn't subscribe to a stream
with non-ASCII characters (failing with a UnicodeEncodeError), as well
as many other potential bugs.
(imported from commit f084a4b4b597b85935655097a7b5a163811c4d71)
This cache should save 2 database queries whenever we send a private
message. However, previously it was per-process (which meant it was
mostly useless) and also buggy (it never stored anything in the cache,
so that it was completely useless). Switching this to our standard
memcached setup will address both problems.
(imported from commit 1d807f30704bccf28de33a80523488aedc58a9be)
Previously we only used these caches for Tornado requests, because we
were not updating memcached when e.g. the user's pointer changed, and
so functions like update_pointer would not work correctly.
Now that we are updated memcached when the User and UserProfile
objects change, we can use these for all requests.
This saves 2 database queries on every Django request to the server.
(imported from commit aa5bffd885d14bde38b95e80a226bd5ab66f253d)
This should substantially decrease the amount of server load generated
by the userpresence system.
I tested that this indeed was indeed saving one query on
/json/update_active_status requests on my laptop with 2 users from the
humbughq.com realm logged in.
(imported from commit 03e9d4eb95b9f664d489862684ae162db2076e08)
This should substantially improve the repeat-rendering time for pages
with large numbers of tweets since we don't need to go all the way to
twitter.com, which can take like a second, to render tweets properly.
To deploy this commit properly, one needs to run
./manage.py createcachetable third_party_api_results
(imported from commit 01b528e61f9dde2ee718bdec0490088907b6017e)
This allows us to handle the return_messages_immediately part of
get_updates requests without having to talk to the database.
(imported from commit ed0b7742d359efb21a0a4960f4fc25f4337e9ad4)
Tim Abbott
Zev Benjamin
Leo Franchi
Luke Faraone
Jessica McKellar
Keegan McAllister