Current implementation uses a caching library called Stash (, via Stash-bundle. If this changes, then the configuration format will most likely change as well.). Stash supports the following cache handlersbackends: FileSystem, Memcache, APC, Sqlite, XCacheRedis and BlackHole.
Use of Memcached or Redis is a requirement for use in Clustering setup, for overview of clustering feature see Clustering.
If eZ Publish Platform changes to another cache system, configuration will change in the future, changes to configuration in StashBundle is listed here:
StashBundle version bundled with 5.4/2014.07 and higher refers to cache backends as "drivers" where it was previously referred to as "handlers" in yml configuration
The cache system is exposed as a "cache" service, and can be reused by any other service as described on the Persistence cache page.
stash: caches: default: # For eZ Publish Platform versions prior to 5.4/2014.07, use "handlers" instead of "drivers"! drivers: - FileSystem inMemory: false registerDoctrineAdapter: false
stash: caches: default: # For eZ Publish Platform versions prior to 5.4/2014.07, use "handlers" instead of "drivers"! drivers: - BlackHole inMemory: true registerDoctrineAdapter: false
If you want to change to another handlercache backend, see in Stash handlers backend configuration below for what kind of settings you have available.
Multi repository setup
New in 5.2 is the possibility to select a specific Stash cache pool on a siteaccess or sitegroup level, the following example shows use in a sitegroup:
ezdemo_group: cache_pool_name: "default" database: ...
The "default" here refers to the name of the cache pool as specified in the stash configuration block shown above, if your install has several repositories (databases), then make sure every group of sites using different repositories also uses a different cache pool to avoid unwanted effects.
NB: We plan to make this more native in the future, so this setting will someday not be needed.
cache backend configuration
To check which cache settings are available for your installation, run the following command in your terminal :
php ezpublish/console config:dump-reference stash
This cache handler backend is using local filesystem, by default the Symfony cache folder, as this is per server, it does not support multi server (cluster) setups!
|We strongly discourage you from storing cache files on NFS, as it defeats the purpose of the cache: speed|
|The path where the cache is placed, default is |
|Number of times the cache key should be split up to avoid having to many files in each folder, default is 2.|
|The permissions of the cache file, default is 0660.|
|The permission of the cache file directories (see dirSplit), default is 0770.|
Limit on how many key to path entries are kept in memory during execution at a time to avoid having to recalculate the path on key lookups, default 200.
|Algorithm used for creating paths, default md5. Use crc32 on Windows to avoid path length issues.|
If you are using a Windows OS, you may encounter an issue regarding long paths for cache directory name. The paths are long because Stash uses md5 to generate unique key that are sanitized really quickly.
Solution is to change the hash algorithm used by Stash.
This configuration is only recommended for Windows users.
FileSystem cache backend troubleshooting
By default, Stash Filesystem cache backend stores cache to a sub-folder named after the environment (
ezpublish/cache/prod). This can lead to the following issue : if different environments are used for operations, persistence cache (manipulating content, mostly) will be affected and cache can become inconsistent.
To prevent this, there are 2 solutions :
Always use the same environment, for web, command line, cronjobs...
Share stash cache across Symfony environments (prod / dev)
Either by using another Stash cache backend, or by setting Stash to use a shared cache folder that does not depend on the environment.
stash: caches: default: FileSystem: path: "%kernel.root_dir%/cache/common"
This will store stash cache to
This cache backend is using shard memory using APC's user cache feature, as this is per server, it does not support multi server (cluster) setups, unless you use PHP-FPM on a dedicated server (this way, user cache can be shared among all the workers in one pool).
As APC user cache is not shared between processes, it is not possible to clear the user cache from CLI, even if you set
Please also note that the default value for
apc.shm_size is 128MB. However, 256MB is recommended for APC to work properly. For more details please refer to the APC configuration manual.
|The time to live of the cache in seconds, default set to 500 (8.3 minutes)|
|A namespace to prefix cache keys with to avoid key conflicts with other eZ Publish sites on same eZ Publish installation, default is |
Stash supports both the php Memcache and Memcached -memcache and php-memcached extensions. However Memcached is recommended as it is much more advanced and supports all settings below. only php-memcache is officially tested on Redhat/Centos while php-memcached is on Debian and Ubuntu. If you have both extensions installed, Stash will automatically choose Memcachedphp-memcached.
Array of Memcached servers, with host/IP, port and weight
|A namespace to prefix cache keys with to avoid key conflicts with other eZ Publish sites on same eZ Publish installation (default is an empty string). |
Must be the same on all server with the same installation. See Memcached prefix_key option. ✲
|default true. See Memcached compression option. ✲|
|default false. See Memcached libketama_compatible option ✲|
|default false. See Memcached buffer_writes option ✲|
|default false. See Memcached binary_protocol option ✲|
|default false. See Memcached no_block option ✲|
|default false. See Memcached tcp_nodelay option ✲|
|default 1000. See Memcached connection_timeout option ✲|
|default 0. See Memcached retry_timeout option ✲|
|default 0. See Memcached send_timeout option ✲|
default 0. See Memcached recv_timeout option ✲
|default 1000. See Memcached poll_timeout option ✲|
|default false. See Memcached cache_lookups option ✲|
|default 0. See PHP Memcached documentation ✲|
|See Memcached socket_send_size option|
|. ✲ ✸|
|See Memcached socket_recv_size option. ✲ ✸|
|See Memcached serializer option. ✲ ✸|
|See Memcached hash option. ✲ ✸|
Specifies the method of distributing item keys to the servers. See Memcached distribution option. ✲ ✸
✲ All settings but servers are only available with memcached php extension, for more information on these settings and which version of php-memcached they are available in, see: http://php.net/Memcached
When using Memcache handler, it's strongly recommended to also use
✸ If you are on eZ Publish 5.1, make sure to update Stash and StashBundle to get access to these settings.
Example with Memcache
stash: caches: default: # For eZ Publish Platform versions prior to 5.4/2014.07, use "handlers" instead of "drivers"! drivers: [ Memcache ] inMemory: true registerDoctrineAdapter: false Memcache: prefix_key: ezdemo_ retry_timeout: 1 servers: - server: 127.0.0.1 port: 11211
If memcached does display connection errors when using the default (ascii) protocol, switching to binary protocol (in the stash configuration and memcached daemon) should resolve the issue.
Memory consumption issues with Redis
If you use the Redis cache driver and encounter problems with high memory consumption, you can use the following (non-SiteAccess-aware) settings:
ezpublish: stash_cache: igbinary: true/false lzf: true/false
ezpublish.stash_cache.igbinaryenables you to use the igbinary PHP extension to serialize objects stored in cache.
- Recommended enabled in most use cases.
ezpublish.stash_cache.lzfenables you to use the LZF PHP extension to compress serialized objects stored in cache.
- Compression consumes some more CPU. A scenario where it improves overall performance is when Redis memory and/or network bandwidth is limited, with spare CPU capacity on PHP server.
These options in combination results in around 1/2 memory usage for API caching compared to not being enabled, igbinary accounts for ~75% of that and LZF the remaining ~25% when configured for max compression.
After changing these settings you need to clear cache and purge Redis content .