We are running a quite large site in production with a several thousand requests per minute. We are running on-prem using multiple Windows servers, one SQL server database and Udp Unicast to synchronize the cache between servers.
For a specific use-case we decided to use the "Content Definitions API" to create and update various content types. These content types can contain from a single property up to >100 properties. When test running this solution we noticed that when, under load, the site stops responding when we POST/PUT a content type to the API. The hickup can last from a couple of seconds to up to a minute. In rare cases the site even stopped responding for such a long period that we had to restart the servers. The errors found in the log afterwards pointed to SQL locks and timeouts.
Looking at the code for creating/updating a content type it's noticeable that every property is persisted separately and each save leads to what seems to be a complete cache wipeout (the log is full of: "WARNING: Cache cleared" messages).
Has anyone experienced similar problems? Is it expected to get this behaviour (maybe it is as the cache is cleared)? How can we make sure that the site wont crash when posting content types via the API that is provided for doing just that (creating content types, not crashing the site ;-))
more information at https://www.optimizely.com/support/ . Please expect to provide profiler traces or copy of your databases for further investigation. I am not part of the support team but I usually assist them in cases like this.
Hi,
We are running a quite large site in production with a several thousand requests per minute. We are running on-prem using multiple Windows servers, one SQL server database and Udp Unicast to synchronize the cache between servers.
For a specific use-case we decided to use the "Content Definitions API" to create and update various content types. These content types can contain from a single property up to >100 properties. When test running this solution we noticed that when, under load, the site stops responding when we POST/PUT a content type to the API. The hickup can last from a couple of seconds to up to a minute. In rare cases the site even stopped responding for such a long period that we had to restart the servers. The errors found in the log afterwards pointed to SQL locks and timeouts.
Looking at the code for creating/updating a content type it's noticeable that every property is persisted separately and each save leads to what seems to be a complete cache wipeout (the log is full of: "WARNING: Cache cleared" messages).
Has anyone experienced similar problems? Is it expected to get this behaviour (maybe it is as the cache is cleared)? How can we make sure that the site wont crash when posting content types via the API that is provided for doing just that (creating content types, not crashing the site ;-))
Thanks in advance!