Nodes

Nodes are simply hosts you have ssh access to that run jobs. Nodes themselves can also be created from jobs for simulating topologies

Jobs

Jobs are the workload that fission is managing and running across multiple nodes for you. this is generally a container workload but may also be simple program

Facts

Facts are used for selecting which nodes are suitable candidates for a particular job, eg some nodes may export a HAS_SSD fact that a high performance disk IO heavy job (eg mysql) may want to select on

Extras

If you are using symlinks to templates that expand on the filename (eg collection of jobs) then there may be occasions where you want to override specific facts (eg code or machine upgrades). normally this would involve updating the config file but as this is symlinked you end up affecting all hosts. the solution for this is a ‘.extras’ file containing additional information that is overplayed on the top of the config to allow further customization allowing you to override a subset or all hosts with information to complete the migration before updating the main template

Tags

tags are used to limit or select a subset of nodes or jobs, eg for listing purposes. tags can also be used for allocation in a similar manner to facts above however the use of facts is recommended. The exception to this rule being geographical placement (ie only allocate on rack37)

if you need to negate a tag, prepending it with ‘-’ can be used to indicate that you want to remove an object from selection if it has this tag. for cli parsing, arguments starting with ‘-’ are interpreted as flags and as such if you only need to specify negation an empty tag can be placed at the beginning of the list to prevent the tag from being interpreted as a cli flag eg “,-physical,-slow”

All jobs are tagged with the ‘job’ tag and all nodes with the ‘node’ tag to make selection easier (eg all nodes but a subset of jobs or a subset of nodes and all jobs). this can also be used with negation to ‘carve away’ unwanted nodes instead of adding them in.

All jobs and nodes are tagged with their hostname prepended with ‘@’ this is to facilitate the selection of individual nodes that may not have a unique tag eg “@mymachine1,@mymachine2,@mymachine30”

Filters

Filters are the counterpart to the tag. These are used by jobs to limit placement decisions. This may be helpful for compliance (credit card processing only on nodes tagged ‘pci-compliant’) or to limit jobs to geographical region

Filters may or may not be used to select based on facts, this is currently undecided due to complexity (ie only on mem > 300MB)