HyperDB causing frequent MySQL connection errors

  • wp_timmiller

    (@wp_timmiller)


    10 years, 10 months ago

    Please read fully. The fifth paragraph indicates why I’m posting in this forum.

    I have a MySQL replication group with a single master and a single slave. I have a farm of three web servers running WordPress and each server is configured with HyperDB to read/write to the MySQL master first and then read from the slave second if the master should be unavailable. This configuration has been running very well for me for over a year.

    Recently (I think it corresponded with my upgrade from WP 3.4.2 to 3.5 or from 3.5 to 3.5.1), I’ve been seeing database connection errors between WP and MySQL. I was running MySQL 5.5.20 and the symptom I was seeing was that servers in the web farm would become blocked from further connections to MySQL because they exceeded the max_connect_errors limit. That parameter has been set to the default of 10 since the cluster was originally configured.

    To be clear, the WordPress on each of the web servers functions normally, until this limit is reached. Then it can’t find the database and redirects to the Install WordPress page. It seems like each page hit is causing at lease one connection error to be recorded. As a short-term fix, I had to increase the max_connect_errors variable to 1000 and regularly run flush hosts to keep MySQL from refusing connections.

    I went ahead and upgraded MySQL to 5.6.12 last weekend so I could get more insight into what was happening with the new performance_schema.host_cache table. While troubleshooting with this guide from MySQL, I decided to temporarily turn off HyperDB on one of the servers in the farm, and the number of connection errors from that server immediately stopped incrementing.

    Incidentally, it’s the Aborted_connects status variable that’s going crazy, and it’s the host_cache.count_handshake_errors field that’s growing for each server in the web farm. No warnings are being logged to the error_log regarding these aborted connections (and yes, I have log-warnings turned on).

    Since I can’t find any other references to this exact problem after much searching, I’m leaning toward it being a problem with my configuration of HyperDB, rather than a bug in the code that was exposed by a WordPress update. However, I’m stumped on where to go next. Has anyone else seen anything similar? Are there any good ideas out there on how to find my problem?

    Thanks,

    Tim

    http://wordpress.org/extend/plugins/hyperdb/

Viewing 2 replies - 1 through 2 (of 2 total)
  • marshalld

    (@marshalld)

    10 years, 7 months ago

    Hi Tim,

    I found the same issue.

    For me, it was cause by the way HyperDB checks for TCP responsiveness.

    I disabled the check and it went away.
    //$wpdb->check_tcp_responsiveness = true;
    $wpdb->check_tcp_responsiveness = false;

    For some reason, MySQL doesn’t like the TCP responsiveness chacks.

    When you look at the stats in MySQL, you can see that the handshake errors go up every time HyperDB does a check.

    mysql> select SUM_CONNECT_ERRORS, COUNT_HANDSHAKE_ERRORS from host_cache where host = ‘back-0’;
    +——————–+————————+
    | SUM_CONNECT_ERRORS | COUNT_HANDSHAKE_ERRORS |
    +——————–+————————+
    | 80 | 80 |
    +——————–+————————+
    1 row in set (0.00 sec)

    Can this issue be escalated to get it resolved?

    JordanThurston

    (@jordanthurston)

    10 years, 1 month ago

    This ^

    I too am getting this error. Constantly.
    If we could get some feedback from devs on specifics and possible solutions I would be most grateful!

Viewing 2 replies - 1 through 2 (of 2 total)
  • The topic ‘HyperDB causing frequent MySQL connection errors’ is closed to new replies.