Memcache monitoring and flush using the nc command

    See more
    • Updated

    Most Drupal website administrators are interested in quickly monitoring memcache's statistics, and the nc command (on some systems, netcat) is a useful command for interacting with memcache servers.

    Find the servers

    As a preliminary step, you'll need a list of the memcache servers that Drupal is using. As an Acquia Cloud customer, you can find your memcache servers by navigating to the Application > Environment > Servers page in the Acquia Cloud UI. Find the "web" server(s) that have "Memcache" and note their full DNS name (e.g. "web-123.prod.hosting.acquia.com").

    If you have a Drupal 6 or Drupal 7 application, you can use the following Drush command to look up the memcache_servers variable:

    drush vget memcache_servers

    The output should look something like this:

    memcache_servers: Array (
  [server-1234.prod.hosting.acquia.com:11211] => default
  [server-5678.prod.hosting.acquia.com:11211] => default
)

    This array is automatically populated on Acquia Cloud servers using memcache.

    Note for Acquia Cloud Free users

    Acquia Cloud Free does not have access to memcache.

    The format of the servers' array is hostname:port => bin. In the sample output, the default memcache bin is split across two memcache hosts — server-1234.prod.hosting.acquia.com and server-5678.prod.hosting.acquia.com, using port 11211 on each host.

    View memcache statistics

    The Memcached protocol provides a stats command that can provide some insight into a few dozen server variables. Many of these server variables count the number of times a particular event occurred since the service was launched.

    In this example, we echo two commands together: stats, followed by quit (separated by a newline). The -q1 option tells the nc command to not disconnect immediately. Without this option, nc might close the connection before it delivers the output.

    /bin/echo -e 'stats\nquit' | nc -q1 [memcache hostname] [port number]

    Here's an example using server information from the previous example:

    /bin/echo -e 'stats\nquit' | nc -q1  $(hostname -s) 11211

    The output of the preceding command should look similar to the following:

    STAT pid 1216 
STAT uptime 2472376 
STAT time 1373572082 
STAT version 1.4.5 
STAT pointer_size 64 
STAT rusage_user 66.790000 
STAT rusage_system 110.900000 
STAT curr_connections 3 
STAT total_connections 297279 
STAT connection_structures 10 
STAT cmd_get 2115470 
STAT cmd_set 92126 
STAT cmd_flush 0 
STAT get_hits 1918776 
STAT get_misses 196694 
STAT delete_misses 806 
STAT delete_hits 1566 
STAT incr_misses 0 
STAT incr_hits 19 
STAT decr_misses 0 
STAT decr_hits 0 
STAT cas_misses 0 
STAT cas_hits 0 
STAT cas_badval 0 
STAT auth_cmds 0 
STAT auth_errors 0 
STAT bytes_read 346432348 
STAT bytes_written 3932785939 
STAT limit_maxbytes 67108864 
STAT accepting_conns 1 
STAT listen_disabled_num 0 
STAT threads 1 
STAT conn_yields 0 
STAT bytes 46704290 
STAT curr_items 21955 
STAT total_items 92127 
STAT evictions 13531 
STAT reclaimed 0 
END

    Monitor memcache statistics

    You can use the watch command to monitor all of these statistics in real time. For example:

    watch -d -n 10 "/bin/echo -e 'stats\nquit' | nc -q1  $(hostname -s) 11211"

    Adding the -d option to the watch command highlights the parts of the output that have changed since the last output. The -n option sets a delay interval; in this example, it's set to 10 seconds.

    Flush memcache

    In most cases, when a particular part of a Drupal page appears "stuck" in cache, you should use the drush cache-clear command. This command can clear a specific cache that appears to be problematic. In extremely rare circumstances you may need to manually trigger a memcache flush. For Drupal 7 websites, you may need to trigger a registry rebuild function. (For more information, see the Registry Rebuild module on drupal.org.) If these processes fail, you can attempt a manual memcache flush. The memcache protocol flush command allows you to immediately flag all stored values as invalid.

    Here's an example:

    /bin/echo -e 'flush_all\nquit' | nc -q1 $(hostname -s) 11211

    If the command completes normally, the output should simply be OK.

    Topics

    Related articles