In local, Tugboat, and BRD environments, Drupal uses Memcache and the memcache PHP module to improve performance by relieving the load on the RDBMS.
BRD utilizes a Memcache cluster on AWS ElastiCache. The PECL memcache module is installed and configured via Ansible and the Memcache nodes are listed explicitly in the CMS_MEMCACHE_NODES environment variable.
This environment variable is read, split, and mapped in settings.brd_common.php to populate $settings['memcache']['servers'].
Tugboat creates a single-node Memcache cluster, available at the hostname memcache and the default port of 11211.
ddev creates a single-node Memcache cluster, available at the hostname memcache and the default port of 11211.
Memcache can be tested and interacted with directly using telnet.
On BRD (using ssm-session):
$ ssm-session vagov-staging cms-test
$ sudo su -
$ source /etc/sysconfig/httpd
$ sudo yum install telnet
...
$ telnet "${CMS_MEMCACHE_NODES%,*}" 11211 # Connect to the first node.On Tugboat:
$ tugboat shell $PREVIEW_ID
$ apt-get install telnet
...
$ telnet memcache 11211Locally:
$ ddev ssh
$ apt-get install telnet
...
$ telnet memcache 11211While telnetted into Memcache, the stats command will return a (long) list of statistics useful for confirming basic health and function of the cluster.
sh-4.2$ telnet memcache 11211
Trying 10.247.34.20...
Connected to 10.247.34.20.
Escape character is '^]'.
stats
STAT pid 1
STAT uptime 1198799
STAT time 1618942692
STAT version 1.6.6
STAT libevent 2.0.21-stable
STAT pointer_size 64
STAT rusage_user 71.834122
STAT rusage_system 73.660193
STAT max_connections 65000
STAT curr_connections 7
STAT total_connections 13695
STAT rejected_connections 0
STAT connection_structures 73
STAT response_obj_bytes 112840
STAT response_obj_total 91
STAT response_obj_free 90
STAT response_obj_oom 0
STAT read_buf_bytes 180224
STAT read_buf_bytes_free 163840
STAT read_buf_oom 0
STAT reserved_fds 10
STAT cmd_get 1395862
STAT cmd_set 135456
STAT cmd_flush 2
STAT cmd_touch 0
STAT cmd_config_get 79914
STAT cmd_config_set 1
STAT cmd_meta 0
STAT get_hits 684482
STAT get_misses 711380
STAT get_expired 0
STAT get_flushed 0
STAT delete_misses 3021
STAT delete_hits 2290
STAT incr_misses 0
STAT incr_hits 0
STAT decr_misses 0
STAT decr_hits 0
STAT cas_misses 0
STAT cas_hits 0
STAT cas_badval 0
STAT touch_hits 0
STAT touch_misses 0
STAT auth_cmds 0
STAT auth_errors 0
STAT bytes_read 503571337
STAT bytes_written 5104566171
STAT limit_maxbytes 13932429312
STAT launch_time_maxbytes 13932429312
STAT accepting_conns 1
STAT listen_disabled_num 0
STAT time_in_listen_disabled_us 0
STAT threads 2
STAT conn_yields 0
STAT hash_power_level 16
STAT hash_bytes 524288
STAT hash_is_expanding 0
STAT curr_config 1
STAT slab_reassign_rescues 0
STAT slab_reassign_chunk_rescues 0
STAT slab_reassign_evictions_nomem 0
STAT slab_reassign_inline_reclaim 0
STAT slab_reassign_busy_items 0
STAT slab_reassign_busy_deletes 0
STAT slab_reassign_running 0
STAT slabs_moved 0
STAT lru_crawler_running 0
STAT lru_crawler_starts 92310
STAT lru_maintainer_juggles 13914112
STAT malloc_fails 0
STAT log_worker_dropped 0
STAT log_worker_written 0
STAT log_watcher_skipped 0
STAT log_watcher_sent 0
STAT bytes 103454004
STAT curr_items 32001
STAT total_items 135456
STAT slab_global_page_pool 0
STAT expired_unfetched 24747
STAT evicted_unfetched 0
STAT evicted_active 0
STAT evictions 0
STAT reclaimed 36231
STAT crawler_reclaimed 0
STAT crawler_items_checked 665253
STAT lrutail_reflocked 771
STAT moves_to_cold 154160
STAT moves_to_warm 75333
STAT moves_within_lru 58759
STAT direct_reclaims 0
STAT lru_bumps_dropped 0
ENDIf Memcache holds corrupted data, or for some other reason should be cleared manually (e.g. if Drush does not work, or for debugging), the flush_all command can be used:
stats
...
STAT bytes 111303
STAT curr_items 27
...
flush_all
OK
stats
...
STAT bytes 0
STAT curr_items 0
...Most of the stats (hits, misses, connections, etc) will be retained, so this is not equivalent to restarting the cluster.
If ElastiCache or Memcache fails or is not available for any reason (networking, PHP extension issues, etc), recovery should be easily accomplished.
The critical code instructing Drupal to use Memcache (rather than just informing Drupal that the cluster exists) is located near the bottom of settings.php:
// Memcache-specific settings
if (extension_loaded('memcache') && !empty($settings['memcache']['servers'])) {
$settings['cache']['default'] = 'cache.backend.memcache';
$settings['memcache']['bins'] = [
'default' => 'default',
];
$settings['container_yamls'][] = $app_root . '/' . $site_path . '/../default/services/services.memcache.yml';
}The conditional depends on two factors, either of which will be sufficient to disable Memcache:
- The PECL Memcache module being loaded and configured for use by PHP.
- The
$settings['memcache']['servers']array being populated.
The PHP ini file that configures Memcache can be located with the following command:
$ php --ini | grep memcache
...Thus, on BRD, the PHP Memcache extension can be disabled temporarily for incident response:
$ mv /etc/opt/remi/php81/php.d/40-memcache.ini ./ # Output of command above
$ systemctl restart httpd
...
$ php -m | grep -i memcache # Should not list the Memcache module
...The Memcache extension will be restored on the next deploy, or the file can simply be moved back into place and Apache reloaded again to re-enable it.
If some profound issue with Memcache or its configuration is discovered, Memcache can be disabled by setting CMS_MEMCACHE_NODES to an empty string or unsetting it completely.
CMS_MEMCACHE_NODES is set in ansible/deployment/config/cms-vagov-prod.yml.
This change will persist across deploys.
On Tugboat, the path to the ini file will differ:
$ mv /usr/local/etc/php/conf.d/docker-php-ext-memcache.ini ./ # Output of command above
$ sudo /etc/init.d/apache2 reload
...
$ php -m # Should not list the Memcache module
...The Memcache extension will be restored on the next deploy, or the file can simply be moved back into place and Apache reloaded again to re-enable it.