Skip to:
Content

BuddyPress.org

Opened 12 years ago

Closed 12 years ago

#4251 closed defect (bug) (duplicate)

Slow queries on Activity Stream during peak times on site.

Reported by: thebeardedone's profile TheBeardedOne Owned by:
Milestone: Priority: normal
Severity: blocker Version: 1.5.5
Component: Core Keywords:
Cc:

Description

We have Buddypress 1.5.5 install on WP 3.3.1 that is going into a tail spin during peak times. The load is all DB related, since we have a dedicated web/app server and DB server. At the time of this report, the number of registered users for the site is between 18000-19000 (we've acquired these number of users in less than a month of being online). I'm attaching a cleaned (removed identifiable info) mysqld.slow-query-log snippet. The number of MySQL slow queries that will get logged are in the thousands, but are basically the same query. The DB server is running Gentoo Linux, with MySQL 5.1.62-r1, and has 32 cores and 64GB of RAM. Due to Buddypress not being compatible with HyperDB, I am unable to mitigate this problem by throwing more DB servers at it.

Attachments (1)

SlowQueryLog-Cleaned.txt (31.5 KB) - added by TheBeardedOne 12 years ago.

Download all attachments as: .zip

Change History (2)

#1 @boonebgorges
12 years ago

  • Milestone Awaiting Review deleted
  • Resolution set to duplicate
  • Status changed from new to closed

Thanks for the log. We've made significant improvements in BP 1.6. See the extended discussion in https://buddypress.trac.wordpress.org/ticket/4045 I encourage you to CC yourself on that ticket and chip in with any additional insights you may have gleaned from your experience.

In the meantime, you're likely to see performance gains by using some sort of persistent object cache, as well as by fine-tuning your MySQL cache.

Due to Buddypress not being compatible with HyperDB, I am unable to mitigate this problem by throwing more DB servers at it.

BuddyPress is definitely compatible with SharDB (and by extension, I believe, with HyperDB), but you're correct that the DB tables themselves cannot be sharded.

Note: See TracTickets for help on using tickets.