The current version should only be used for testing; for example,
if you want to create a bunch of streams for stress testing, you
can run this in a loop.
(imported from commit ec51a431fb9679fc18379e4c6ecdba66bc75a395)
We need to run the schema migration manually using
"CREATE INDEX CONCURRENTLY upper_stream_name_idx ON zerver_stream ((upper(name)));"
since we need CONCURRENTLY and I seem to recall that doesn't work with South.
This significantly improves the uncached performance of get_stream()
(e.g. from 32ms to 9ms). At present, this codepath is not used
particularly heavily since we do cache the stream names and do most of
our filtering by recipient ID, but the index isn't expensive and does
provide a significant improvement in the uncached case.
(imported from commit 4d28dc2e9a02d0602861b165393d90ed18f5f4c8)
We need to run the schema migration manually using
"CREATE INDEX CONCURRENTLY upper_subject_idx ON zerver_message ((upper(subject)));"
since we need CONCURRENTLY and I seem to recall that doesn't work with South.
Apparently our existing indexes on subject/topic weren't being used in
our narrowing queries, because we do case-insensitive search.
This substantially improves our database performance around
stream+topic narrows. See before and after query plans below from my
test instance.
humbug=# explain analyze SELECT "zerver_message"."id" FROM "zerver_message" WHERE ("zerver_message"."recipient_id" = 38 AND UPPER(zerver_message.subject) = 'TEST' AND "zerver_message"."id" <= 348495 ) ORDER BY "zerver_message"."id" DESC LIMIT 50;
QUERY PLAN
---------------------------------------------------------------------------------------------------------------------------------------------------------
Limit (cost=13510.61..13510.71 rows=41 width=4) (actual time=32.952..32.958 rows=2 loops=1)
-> Sort (cost=13510.61..13510.71 rows=41 width=4) (actual time=32.946..32.947 rows=2 loops=1)
Sort Key: id
Sort Method: quicksort Memory: 25kB
-> Bitmap Heap Scan on zerver_message (cost=237.99..13509.51 rows=41 width=4) (actual time=2.357..32.912 rows=2 loops=1)
Recheck Cond: (recipient_id = 38)
Filter: ((id <= 348495) AND (upper((subject)::text) = 'TEST'::text))
-> Bitmap Index Scan on zephyr_message_recipient_id (cost=0.00..237.98 rows=8221 width=0) (actual time=1.178..1.178 rows=10354 loops=1)
Index Cond: (recipient_id = 38)
Total runtime: 33.049 ms
(10 rows)
humbug=# explain analyze SELECT "zerver_message"."id" FROM "zerver_message" WHERE ("zerver_message"."recipient_id" = 38 AND UPPER(zerver_message.subject) = 'TEST' AND "zerver_message"."id" <= 348495 ) ORDER BY "zerver_message"."id" DESC LIMIT 50;
QUERY PLAN
---------------------------------------------------------------------------------------------------------------------------------------------------------------
Limit (cost=435.11..435.22 rows=41 width=4) (actual time=4.998..4.999 rows=2 loops=1)
-> Sort (cost=435.11..435.22 rows=41 width=4) (actual time=4.997..4.997 rows=2 loops=1)
Sort Key: id
Sort Method: quicksort Memory: 25kB
-> Bitmap Heap Scan on zerver_message (cost=275.63..434.02 rows=41 width=4) (actual time=4.981..4.984 rows=2 loops=1)
Recheck Cond: ((upper((subject)::text) = 'TEST'::text) AND (recipient_id = 38))
Filter: (id <= 348495)
-> BitmapAnd (cost=275.63..275.63 rows=41 width=0) (actual time=4.954..4.954 rows=0 loops=1)
-> Bitmap Index Scan on upper_subject_idx (cost=0.00..37.38 rows=1744 width=0) (actual time=2.972..2.972 rows=27457 loops=1)
Index Cond: (upper((subject)::text) = 'TEST'::text)
-> Bitmap Index Scan on zephyr_message_recipient_id (cost=0.00..237.98 rows=8221 width=0) (actual time=0.855..0.855 rows=10354 loops=1)
Index Cond: (recipient_id = 38)
Total runtime: 5.049 ms
(13 rows)
(imported from commit 1f4815ccb0691053ff8d505149482dbc74153fb3)
Don't warn when @-mentioning a bot on a public stream that it does
not appear to be subscribed to. It may be receiving those messages
anyway.
(imported from commit 4a00694942a721897a01736f48033c71048e0b16)
It makes the event queue return all messages on public streams, rather
than only the user's subscriptions. It's meant for use with chat bots.
(imported from commit 12d7e9e9586369efa7e7ff9eb060f25360327f71)
This doesn't address the more complicated case of someone @-mentioning
you on a muted topic, which consensus is you do want to get
information for, but we need to develop some infrastructure to present
that case to users clearly.
(imported from commit a4bc1e89c108fa8ba6eccc0a198eabf2231326ab)
By far the common case for get_old_messages is the home view loading
queries, for which we have raw queries. This patch substantially
improves those queries using the observation that we weren't actually
using the zerver_message table that we were joining with.
I actually expect this to result in a noticable performance
improvement for loading of the homepage.
(imported from commit 12807e5a74eb63275b2523a5f62fd901ab632f0f)
Deployment instructions: I think all the queue workers get
restarted automatically, so there is probably nothing special
to do here in the deploy itself, but we will want to monitor
it closely, and the change should make our number of locks go
down.
QueueProcessingWorker.start() now calls consume_and_commit(),
which ensures that we don't hold locks after work actions
by using Django's commit_on_success() decorator.
Obviously, workers that override start() will not call consume_and_commit()
through this code path. SlowQueryWorker calls commit_on_success()
in its start() method now, and I hope to address MissedMessageWorker soon.
(imported from commit f3f38a7f45730eee8f3b5794371ba5b994017676)
These are some queries on API usage, desktop usage, and
Android usage that would be of interest to Waseem. These
will eventually be subsumed into /activity, but some interim
data issues may make them easier to keep separate for now.
(imported from commit 697a8496cbf4447d557a3fc89f64c1c4d3e67e70)
These parameters collectively determine when very old transaction ids
are replaced with the special FrozenXID transaction id, which is
older than any other transaction id. Old transaction ids must be
periodically replaced like this to prevent transaction id wraparound.
The only disadvantage of increasing autovacuum_freeze_max_age is
increased disk usage, but a value of 2 billion should only require
500MB, which is pretty trivial.
Since changing autovacuum_freeze_max_age requires a restart, we will
still have to issue a manual VACUUM, with vacuum_freeze_min_age and
vacuum_freeze_table_age temporarily set to lower values.
(imported from commit 22f9ecdfc5b6a07918771d541192aa6d6369878a)
The other zulip.net check was for a transition to using that file that
has now completed.
(imported from commit 991d9165515b5611865957255f9da7a69a75fd7b)
In order to support iOS Push Notifications, we need to keep track
of a device's unique APNS Token. These are delivered to our iOS
code after registering for remote notifications
(imported from commit bbe34483e1380dc20a1c93e3ffa1fcfdb9087e67)
Use the commit_on_success() context manager around the call
to internal_send_message() inside of SlowQueryWorker's polling
loop, so that the pending SELECT statement from
get_status_dict_by_realm() gets committed. If we don't do
this, postgres will hold locks on zerver_userprofile, and other
tables, for a long time, which can interfere with migrations.
This is an interim solution until we switch postgres's default
commit behavior. Right now the default transaction isolation
is "read committed," so SELECT statements lead to AccessShareLocks
that do no get closed until the transaction finishes.
(imported from commit f72aeffbbe71a731e327459f15bd7dbebaf9e0b8)
Trac #1162
The process_fence method replaces code blocks with placeholders, so
indexes stored before the replacement are incorrect. However, because
the closed code blocks have been replaced, we can simply search the
whole string for any remaining opening code block markers.
(imported from commit 6a9e6924840f8f3ca5175da7c52a905e27c1fabd)
The psql command was previously failing because it was trying to use
the non-existant DB user "zulip".
(imported from commit 68ad5c979ce33cc54e9d55f1822cba9fac648944)
I added filter() statements to do_update_message_flags().
Here is some context:
Steve Howell: Case 1, have AND clause to reduce work for DB.
humbug=> update zerver_usermessage set flags = (flags & ~1) where id > 9000;
UPDATE 382
humbug=> select count(*) from zerver_usermessage where (flags & 1) = 0;
count
-------
382
(1 row)
humbug=> explain analyze update zerver_usermessage set flags = (flags | 1) where (flags & 1) = 0;
QUERY PLAN
------------------------------------------------------------------------------------------------------------------------
Update on zerver_usermessage (cost=0.00..266.85 rows=47 width=27) (actual time=5.727..5.727 rows=0 loops=1)
-> Seq Scan on zerver_usermessage (cost=0.00..266.85 rows=47 width=27) (actual time=0.045..2.751 rows=382 loops=1)
Filter: ((flags & 1::bigint) = 0)
Rows Removed by Filter: 9000
Total runtime: 5.759 ms
(5 rows)
humbug=> select count(*) from zerver_usermessage where (flags & 1) = 0;
count
-------
0
(1 row)
Leo Franchi: Sounds reasonable, but I know way less than zev about DBs so I'll defer to his judgement :)
Steve Howell: Case 2, how the code works now:
humbug=> update zerver_usermessage set flags = (flags & ~1) where id > 9000;
UPDATE 382
humbug=> select count(*) from zerver_usermessage where (flags & 1) = 0;
count
-------
382
(1 row)
humbug=> explain analyze update zerver_usermessage set flags = (flags | 1);
QUERY PLAN
---------------------------------------------------------------------------------------------------------------------------
Update on zerver_usermessage (cost=0.00..243.28 rows=9382 width=27) (actual time=362.075..362.075 rows=0 loops=1)
-> Seq Scan on zerver_usermessage (cost=0.00..243.28 rows=9382 width=27) (actual time=0.008..6.138 rows=9382 loops=1)
Total runtime: 362.105 ms
(3 rows)
humbug=> select count(*) from zerver_usermessage where (flags & 1) = 0;
count
-------
0
(1 row)
Steve Howell: In both trials, we set it up so that only 382 of 9382 rows need to be updated. The first trial runs about 63x as fast. The second trial, if my theory is correct, is doing 24x as many writes as it needs. Both trials are reading all 9382 rows.
Steve Howell: The expense of the update statement seems to be proportional to the number of rows you "update", not the number of rows that you actually change.
Steve Howell: For now I created #1869.
Zev Benjamin: That sounds like a reasonable explanation. The disk IO can be expensive
(imported from commit d9090daee1f81cad76c430de0956f9bd504da075)
I looked around the Django docs for a more future-proof way of
addressing this problem of wanting both /apps and /apps/ to work. It
doesn't seem like there is.
Note that APPEND_SLASH is default True, so we're getting a / appended
automatically if there's no URL match without the /:
https://docs.djangoproject.com/en/dev/ref/settings/#append-slashhttp://stackoverflow.com/a/11690144 points out making the slash
optional in the URL regex, but a commenter points out that this is bad
for SEO.
(imported from commit 8e883fb786c583fe5f561f14473211f6b100cecf)
Steve Howell
Tim Abbott
Kevin Mehall
Jessica McKellar
Zev Benjamin
Leo Franchi
Waseem Daher