By now, our servers are configured trough a YAML based Hiera backend for Puppet. Depending on your access rights you are able to modify the whole configuration or only parts of the hierarchy such as domains or single servers.
- common values used on all servers
- here, you add values like global SSH keys or firewall rules
- common values used on all servers in this domain
- here, you add domain specific values like dns resolver for a particular subnet
- values used on this explicit server
- here, you add server specific values such as websites
We will provide you access to the appropriate GIT repository to alter your configuration. Depending on your role you can add global or domain/server specific configurations.
This is a temporary solution only (see below)
We are working hard on a new, HTTP/REST based backend for Hiera. You will be able to alter server specific configurations such as websites trough this new API. In a first step, the API will be backed by the great Swagger UI, later on we will develop our own management interfaces which interacts with this API.
To configure certain, memory-aware services, suche as memcached or
Varnish, we use the so called
memory_ratio parameter to calculate
the amount of memory for a given service.
As basis, we take 80% of the servers total memory. This amount is then
distributed trough the
memory_ratio for a specific service.
Note: You can also use floating point numbers here. e.g.
memory_ratio: 1.2 results in 66% (0.8/1.2)
single Tomcat instance¶
- Server with 4GB Memory
- Tomcat Service, no others
- Result: 4GB Memory * 80% = 3.2GB / 1 (ratio) = 3.2GB Memory for Tomcat
Tomcat instance among others¶
- Server with 4GB Memory
- Website Service, Tomcat Service
memory_ratiomanually set to
- Result: 4GB Memory * 80% = 3.2GB / 3 (ratio) = 1.07GB Memory for Tomcat
Every new server comes with 1GB RAM, 1CPU Core and 20GB Space (regarding to our offers ) This is a very basic setup and it’s recommended to plan and check the sizing of your server carefully before going live. You can simply scale every option of your server - regarding to your actual needs (without paying for unused resources).
Hint: use load tests to plan your sizing. We’re also happy to help you out.
Please plan your architecture / server setup carefully. Normaly we recommend to split services across different servers.
- large TYPO3 website
- Solr search server
- Memcache caching backend
Setup at least 3 server and go live:
- Server 1: Websites with TYPO3 (running Nginx, PHP, MySQL)
- Server 2: Apache Solr
- Server 3: Memcached
Notional: the website went live and now the users love your search function and drop tausends of search querys every second. So now you can easly scale your Solr server without outages of the main website. (because just the search function is overloaded) If you had all services on one server, probably your website is down now, because Solr catched all the available resources.