a9s LogMe2
Fluentd
Buffers
Traditionally, Fluentd has used memory buffers for buffering data. While memory buffers are fast and efficient, they come with the risk of losing data in case of a system failure. In order to mitigate this risk, we have decided to switch to file buffers.
Platform Plan
Since a Platform Operator possibly collects a lot of log messages from different sources, like service instances and framework components, the default cluster template will eventually not suffice. To meet these special requirements from a Platform Operator, it is possible to use the cluster template to create a special service plan that can be used for this purpose.
Normally only one instance with such a plan should exist in a platform.
Scale OpenSearch Nodes
The following changes to the cluster template should only be made in the context of a specific Platform Operator plan and are not recommended for Application Developer plans.
Since it can be very useful to scale the number of OpenSearch nodes based on the load of the platform, the Platform
Operator can use the default placeholder value instances_os of
the cluster template to scale the OpenSearch nodes. If it is not explicitly set, 3 nodes will be used. Although it is
possible to increase the number of OpenSearch nodes to over 10, this is not recommended.
Scale Fluentd Nodes
The following changes to the cluster template should only be made in the context of a specific Platform Operator plan and are not recommended for Application Developer plans.
It can be very useful to scale up the number of Fluentd nodes, as during an update no logs will be streamed when having
only one Fluentd node, which makes debugging a failed update difficult. The Platform Operator can scale up the Fluentd
nodes using the
default placeholder value
instances_fluentd of the cluster template. If it is not explicitly set, only one node will be used.