Improve user in-guild sync process #165
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
ChrisLovering
force-pushed
the
in_guild-sync-perf-improvement
branch
from
05e7053 to
b8a6198
Compare
|
Shouldn't the condition be `if user.id _not_ in guild_member_ids`, since
`guild_member_ids` contains the users that are on the guild? From my
understanding this currently sets `in_guild=False` for all users that
are on the guild (and found in the queryset).
Also, nitpicking but it would be nice to remove the space before `%d` in
the `log.info` line.
|
ChrisLovering
force-pushed
the
in_guild-sync-perf-improvement
branch
from
c298d9a to
e47b967
Compare
Previously we set all users in_guild to False, and relied on users being set back to in_guild when iterating through guild.members However, this caused two problems 1. For a short window a users in_guild status was incorrect 2. It required an update for all users in_guild to be sent to postgres to update in_guild back to True. This diff changes that, so instead only users who are not found in the guild have in_guild set to False. The bottleneck for this query is the number of users that are currently in_guild=False. Testing locally, with 360k users off guild, this took 7.4s to query out, and 0.5s to process & 15.1 s to commit.
ChrisLovering
force-pushed
the
in_guild-sync-perf-improvement
branch
from
e47b967 to
c2fe170
Compare
Yeah, should have marked this as draft as there were a few more things I wanted to do. I've pushed some new commits and updated the PR description with perf stats. |
This means the other, larger, columns do not need to be deserialised, bringing the query time down from 7.4s down to 3.5s. I couldn't simply do select(models.User.id) here, as we need the full User object in order to mutate and update it later inthe process.
ChrisLovering
force-pushed
the
in_guild-sync-perf-improvement
branch
from
c2fe170 to
db22e38
Compare
|
Thanks, this looks good now!
|
jb3
approved these changes
Jul 1, 2024
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Closes #157
Previously we set all users in_guild to False, and relied on users being set back to in_guild when iterating through guild.members
However, this caused two problems
This diff changes that, so instead only users who are not found in the guild have in_guild set to False.
The bottleneck for this process currently is the number of users that need to be committed back to the database.
Testing locally, with 380k users off guild, the perf logging outputs:
in_guild sync: total time 19.140096s, query 3.486625s, processing 0.540042s, commit 15.113428s. The commit timing isn't reliabale, as it is dependent on how long it takes for control to be passed back after awaiting