diff options
Diffstat (limited to 'develop/print.html')
| -rw-r--r-- | develop/print.html | 38 |
1 files changed, 37 insertions, 1 deletions
diff --git a/develop/print.html b/develop/print.html index 417aed049a..584b5b0eb3 100644 --- a/develop/print.html +++ b/develop/print.html @@ -3920,14 +3920,32 @@ with intermittent connections, at the cost of higher memory usage. By default, this is zero, which means that sync responses are not cached at all.</p> </li> +<li> +<p><code>cache_autotuning</code> and its sub-options <code>max_cache_memory_usage</code>, <code>target_cache_memory_usage</code>, and +<code>min_cache_ttl</code> work in conjunction with each other to maintain a balance between cache memory +usage and cache entry availability. You must be using <a href="https://github.com/matrix-org/synapse#help-synapse-is-slow-and-eats-all-my-ramcpu">jemalloc</a> +to utilize this option, and all three of the options must be specified for this feature to work.</p> +<ul> +<li><code>max_cache_memory_usage</code> sets a ceiling on how much memory the cache can use before caches begin to be continuously evicted. +They will continue to be evicted until the memory usage drops below the <code>target_memory_usage</code>, set in +the flag below, or until the <code>min_cache_ttl</code> is hit.</li> +<li><code>target_memory_usage</code> sets a rough target for the desired memory usage of the caches.</li> +<li><code>min_cache_ttl</code> sets a limit under which newer cache entries are not evicted and is only applied when +caches are actively being evicted/<code>max_cache_memory_usage</code> has been exceeded. This is to protect hot caches +from being emptied while Synapse is evicting due to memory.</li> +</ul> +</li> </ul> <p>Example configuration:</p> <pre><code class="language-yaml">caches: global_factor: 1.0 per_cache_factors: get_users_who_share_room_with_user: 2.0 - expire_caches: false sync_response_cache_duration: 2m + cache_autotuning: + max_cache_memory_usage: 1024M + target_cache_memory_usage: 758M + min_cache_ttl: 5m </code></pre> <h3 id="reloading-cache-factors"><a class="header" href="#reloading-cache-factors">Reloading cache factors</a></h3> <p>The cache factors (i.e. <code>caches.global_factor</code> and <code>caches.per_cache_factors</code>) may be reloaded at any time by sending a @@ -6792,6 +6810,24 @@ caches: # #cache_entry_ttl: 30m + # This flag enables cache autotuning, and is further specified by the sub-options `max_cache_memory_usage`, + # `target_cache_memory_usage`, `min_cache_ttl`. These flags work in conjunction with each other to maintain + # a balance between cache memory usage and cache entry availability. You must be using jemalloc to utilize + # this option, and all three of the options must be specified for this feature to work. + #cache_autotuning: + # This flag sets a ceiling on much memory the cache can use before caches begin to be continuously evicted. + # They will continue to be evicted until the memory usage drops below the `target_memory_usage`, set in + # the flag below, or until the `min_cache_ttl` is hit. + #max_cache_memory_usage: 1024M + + # This flag sets a rough target for the desired memory usage of the caches. + #target_cache_memory_usage: 758M + + # 'min_cache_ttl` sets a limit under which newer cache entries are not evicted and is only applied when + # caches are actively being evicted/`max_cache_memory_usage` has been exceeded. This is to protect hot caches + # from being emptied while Synapse is evicting due to memory. + #min_cache_ttl: 5m + # Controls how long the results of a /sync request are cached for after # a successful response is returned. A higher duration can help clients with # intermittent connections, at the cost of higher memory usage. |