Rsyslog """"""" Diagram ''''''' .. graphviz:: digraph "arch" { nodesep=0.6; ranksep=1.7; subgraph cluster_client { label="Client"; "client"[label="rsyslog"]; } subgraph cluster_mgmtclient { label="Management node"; "client2"[label="rsyslog"]; } subgraph cluster_server1 { label="Islet Hypervisor 1"; "server1"[label="rsyslog"]; } subgraph cluster_server3 { label="Worker Hypervisor"; "server3"[label="rsyslog"]; } "gluster"[style=dotted]; client -> server1 [color=blue]; client2 -> server3 [color=green]; server1 -> gluster[color=red]; server3 -> gluster[color=red]; { rank = sink; Legend [shape=none, margin=0, label=<
Legend
Client access
Mgmt client access
File writes
>]; } } Description ''''''''''' The proposed solution for logs handling is to use independant rsyslog servers writing to the `Gluster` filesystem. A unique rsyslog server will be deployed for each islet as well as for management nodes. In case of islet hypervisor(s) failure(s), it will be necessary to manually switch to other islet rsyslog servers. The switch will be documented. Also, there will be a need to configure a rate limit either on the client or on the server side in order to avoid log explosion. This rate limit threshold has to be determined. The tree structure in which log files will be written will be based on client node names and will also take into account the node type (node, switch, etc...). Log messages format will be the same accross all rsyslog servers and shared accross all `Ocean` based servers. If there is a need to export logs to the outside, the :term:`ETL` service can be used. .. raw:: latex \clearpage