Configuration

Note

This document describes the configuration of Sysinspect system.

Sysinspect can run in two modes:

Distributed Mode using network-connected Minions, allowing many “boxes” to be subscribed to a Master command center. This is a typical use.
Solo Mode, i.e. locally, only on the current local “box”, affecting nothing else. This usage is for a very small embedded systems only.

Config Location

Configuration files can be in three locations and are searched in the following order:

Current directory from which the app was launched: sysinspect.conf.
“Dot-file” in the current user’s home ~/.sysinspect
As /etc/sysinspect/sysinspect.conf.

Synopsis

Configuration is located under config section in a YAML file. This section has two important sub-sections:

master for all settings of Sysinspect Master
minion for covering settings of Sysinspect Minion

Config Section

Main section of the entire configuration is config. It is located at the root of the configuration file and contains the following directives:

modules

Path to location of the modules, used in the model and states. Default value is /usr/share/sysinspect/modules according to the LSB standard.

master

Sysinspect Master configuration.

minion

Sysinspect Minion configuration.

Common Configuration

The following configuration is applicable to both Master and Minion in their respective configuration files.

`log.stream`

Type: string

Path to the log stream file. This file is where Sysinspect Master or Minion writes down everything it does—like keeping a diary of its actions and events. If something goes wrong or you want to check what happened, you can look here. The Master or Minion tries to save this file in a few places, depending on what it’s allowed to access:

/var/log/ typically a standard place for logs, but might not be accessible

$HOME/.local/ in user home directory

/tmp/ if anything else fails

Default filename for Master is sysmaster.standard.log and for Minion is sysminion.standard.log.

`log.errors`

Type: string

Path to the log errors file. This file is used to collect all error logs from the Sysinspect Master or Minion. Depends on permissions, the following order is used:

/var/log/ typically a standard place for logs, but might not be accessible

$HOME/.local/ in user home directory

/tmp/ if anything else fails

Default filename for Master is sysmaster.errors.log and for Minion is sysminion.errors.log.

`pidfile`

Type: string

Path to the PID file. This file is used to store the process ID of the Sysinspect Master or Minion. It is important for managing the lifecycle of the service, allowing for proper start/stop operations.

By default, pidfile is located at standard location: /run/user/<UID>/sysmaster.pid for Master and /run/user/<UID>/sysminion.pid for Minion.

Note

Relocate the PID file to a different location only if it is really necessary: e.g. your system is not standard, different permissions etc. Otherwise it is highly recommended to keep the default location as is.

Master

Sysinspect Master configuration is located under earlier mentioned master section, and contains the following directives:

`socket`

Type: string

Path to the local Unix socket that the sysinspect CLI uses to talk to the running Sysinspect Master process.

In practice, this is the control socket on the master host itself. When you run a command such as sysinspect apply ..., the CLI writes the request to this socket and the master daemon forwards the work to minions over the network. This setting does not change the network listener for minions; that is controlled by bind.ip and bind.port.

Default value is /var/run/sysinspect-master.socket.

Change this value only if you need the socket in a different location, for example because /var/run is not writable in your environment or you want to keep runtime files under another service directory. If you change it, make sure both the master service and the sysinspect command use the same configuration file, otherwise the CLI will not be able to reach the master.

`console.bind.ip`

Type: string

IPv4 address on which sysmaster listens for console connections from sysinspect. This is the active command transport between sysinspect and sysmaster.

When this value is 0.0.0.0, the local sysinspect client still connects through 127.0.0.1.

If omitted, the default value is 127.0.0.1.

`console.bind.port`

Type: integer

TCP port for the master’s console endpoint used by sysinspect.

If omitted, the default value is 4203.

`console` key material

The console endpoint uses the master’s RSA material under the master root directory:

console.rsa — console private key

console.rsa.pub — console public key

console-keys/ — authorised client public keys

These are filesystem conventions under the master root, not YAML configuration directives.

`bind.ip`

Type: string

IPv4 address on which the Master is listening for all incoming and outgoing traffic with Minion communication.

Default value is 0.0.0.0.

`bind.port`

Type: integer

Network port number on which the Master is listening using bind.ip directive.

Sysinspect Master port is 4200.

`cmdb-update`

Type: duration

Age threshold for master-side startup inventory / CMDB reconciliation.

On master startup, Sysinspect may refresh stored observed host facts from current minion traits when the CMDB record is older than this threshold. This keeps hostname, FQDN, and IP facts fresh without replacing the minion identity or its stored startup backend.

Default is 1w.

`hopstart.*`

Type: flat dotted keys

Settings for the SSH-backed startup backend used by sysinspect cluster --hopstart.

Hopstart is meant for targets where you cannot or do not want to integrate with the target init system. Typical cases are read-only systems, appliances, embedded boxes, Android-like hosts, or deployments where only one writable application tree exists.

Example:
config:
  master:
    hopstart.batch: 10
    hopstart.network.forward: false
    hopstart.on-start: false

`hopstart.batch`

Type: integer

Maximum number of concurrent hopstart SSH launch attempts.

Default is 10.

`hopstart.network.forward`

Type: boolean

Reserved for future forwarding of hopstart requests through another master or proxy tier.

Default is false.

`hopstart.on-start`

Type: boolean

Reserved for future behavior where master startup can automatically issue hopstart for eligible minions.

Default is false.

Important

Master runs a File Server service. This service is very important for all the minions, as they are exchanging data with the master, by downloading all the required artefacts to be processed on their targets.

File Server service serves static data, which is continuously checked by each minion and updated, if that data changes. In particular, the artefacts are modules, trait configs, models, states etc. Typically, File Server service has the root of all the data in /etc/sysinspect/data.

Warning

Even though as of current version, there is no specific layout of the static data on the File Server service to manager all the artifacts. However, this is a subject to change.

Within the “/data” directory, currently one is free to organise the layout as they want. However, it is strongly advised to keep all the models, states and other artefacts separated from each other, using their own directories and namespaces. Future releases will have configurable default namespaces for each cathegory of the artefacts.

Below are directives for the configuration of the File Server service:

`fileserver.bind.ip`

Type: string

Same as bind.ip, but for the internal File Server service.

`fileserver.bind.port`

Type: integer

Network port number on which the File Server service is listening.

File Server service port is 4201.

`fileserver.models.root`

Type: string

Relative path where are the master models kept. Default: /models.

`fileserver.models`

Type: list

List of subdirectories within fileserver.models, exporting models. If a model is not in the list, it will not be available for the minions.

`fileserver.sensors.root`

Type: string

Relative path where are the master sensors kept. Default: /sensors.

`fileserver.sensors`

Type: list

List of subdirectories within fileserver.sensors, exporting sensors. If a sensor is not in the list, it will not be available for the minions.

`api.enabled`

Type: boolean

Enable or disable the embedded Web API listener inside the sysmaster process so Sysinspect Master can be controlled remotely.

This listener is part of sysmaster itself. Sysinspect does not start a separate Web API daemon for this interface.

Important

When enabled, the WebAPI serves its OpenAPI documentation through Swagger UI. The documentation endpoint is available at https://<HOST>:<PORT>/doc/.

If api.enabled is true but TLS is not configured correctly, sysmaster keeps running and the Web API stays disabled with an error log.

The Swagger UI is typically available at https://<HOST>:<PORT>/doc/. Default port is 4202.

Note

Swagger UI is served whenever the WebAPI is enabled and api.doc is true.

Default is true.

`api.doc`

Type: boolean

Enable or disable the embedded Web API documentation endpoints served by Swagger UI.

When true, the documentation endpoint is available at https://<HOST>:<PORT>/doc/ on the same HTTPS listener as the Web API.

When false, Sysinspect keeps the Web API itself enabled but does not expose the Swagger UI or OpenAPI document endpoint.

Typical usage:
config:
  master:
    api.enabled: true
    api.doc: true
    api.tls.enabled: true
    api.tls.cert-file: /etc/sysinspect/webapi/server.crt
    api.tls.key-file: /etc/sysinspect/webapi/server.key
To require trusted client certificates for the Web API and its documentation endpoints, add a CA bundle:
config:
  master:
    api.enabled: true
    api.doc: true
    api.tls.enabled: true
    api.tls.cert-file: /etc/sysinspect/webapi/server.crt
    api.tls.key-file: /etc/sysinspect/webapi/server.key
    api.tls.ca-file: /etc/sysinspect/webapi/clients-ca.pem
Default is true.

`api.bind.ip`

Type: string

IPv4 address on which the embedded Web API listener accepts traffic.

Default value is 0.0.0.0.

`api.bind.port`

Type: integer

Network port number on which the embedded Web API listener is listening.

The embedded Web API listener uses port 4202 by default.

`api.auth`

Type: string

Authentication method to be used for the embedded Web API. This is a string and can be one of the following:

pam

ldap (planned, not implemented yet)

`api.devmode`

Type: boolean

Enable or disable development-only helpers for the embedded Web API.

Danger

This option is exclusively for development purposes. If enabled, the authentication endpoint returns a static token and the development query helpers remain available. Do not combine api.devmode: true with production exposure of the Web API documentation. endpoint is exposed.

Default is false.

`api.tls.enabled`

Type: boolean

Enable TLS for the embedded Web API listener.

Default is false.

`api.tls.cert-file`

Type: string

Path to the PEM certificate chain used by the Web API TLS listener.

If the path is relative, it is resolved under the Sysinspect root. If it is absolute, it is used as-is.

When api.tls.enabled is true, this option is required.

`api.tls.key-file`

Type: string

Path to the PEM private key used by the Web API TLS listener.

If the path is relative, it is resolved under the Sysinspect root. If it is absolute, it is used as-is.

When api.tls.enabled is true, this option is required.

`api.tls.ca-file`

Type: string

Optional CA bundle path used to verify client certificates for the Web API TLS listener.

If the path is relative, it is resolved under the Sysinspect root. If it is absolute, it is used as-is.

When set, clients must present a certificate chain that validates against this CA bundle.

`api.tls.allow-insecure`

Type: boolean

Allow the embedded Web API to start with a self-signed or otherwise non-public TLS certificate.

When this option is false, Sysinspect rejects a self-signed Web API certificate during startup.

When this option is true, Sysinspect allows that setup and logs a warning so operators know clients must explicitly trust the certificate.

Default is false.

`telemetry.location`

Type: string

Location of the telemetry local database (do not mix with the OTEL or OTEL collector). This is a directory, where the key/value database is located and records all results, coming from the minion when processing a given query. Default is set to /var/tmp/sysinspect/telemetry.

`telemetry`

Type: key/value

The following keys are supported:

collector.grpc
Type: string

This is the location of the telemetry collector. It is a string in format <IP>:<PORT>. This is the location of the telemetry collector, which is used to send all telemetry data to. This is a string and can be one of the following: URI of the telemetry collector in format <IP>:<PORT>. Default value is 127.0.0.1:4317 assuming that the collector is running on the same machine.

collector.compression
Type: string

Compression algorithm to be used for the telemetry collector. This is a string and can be one of the following:

gzip (default)

zstd

none

Which algorithm to choose?

gzip is a good choice for most of the cases. It is most backward compatible but it does not have a good compression ratio and is using more CPU power. On large metric, ratio is about 33.8 and throughput is about 131 MB/s, resulting to about 52K ns/op.

zstd is a much better choice for the embedded systems, where the CPU power is limited. It has a better compression ratio, and is also faster than gzip, but is too new. On large metric, ratio is about 47.2 and throughput is about 476 MB/s, resulting to about 14K ns/op.

none no compression at all. This is a good choice for the embedded systems, where the CPU power is limited and the network bandwidth is not an issue.

Attention

The compression algorithm must be supported by the telemetry collector. Far not all collectors supports zstd compression algorithm.

exporter-resources
Type: key/value

This is a key/value pair, which is used to set the resources for the telemetry exporter. This element can contain any kind of static values. A resource describes the entity producing telemetry (e.g. a process, container, or service). It is immutable and shared by all signals (traces, metrics, logs) coming from that entity.

The default included keys are the following:

service.name (string) — name of the telemetry service. Default value is sysinspect.

service.namespace (string) — namespace of the telemetry service. Default value is sysinspect.

service.version (string) — version of the telemetry service. Default value is the current SysInspect version.

host.name (string) — name of the host. Default value is the hostname of the machine.

os.type (string) — type of the operating system. Default value is linux.

deployment.environment (string) — deployment environment of the operating system. Default value is production.

os.version (string) — version of the operating system. Default value is the current OS version.

Attention

To turn off a specific resource from being exported (redefined or default), not specifying it will lead to a default value. In order to explicitly disable a default resource, set the value to false.

exporter-scope
Type: key/value

This is a key/value pair, which is used to set the queue for the telemetry exporter. Scope are typically a name, version, schema_url attributes etc. The attributes here are optional, build-time metadata that further qualify the scope—e.g. the URL of the instrumentation’s repo, a feature-flag, or other static info about that library.

The default included keys are the following:

name — name of the scope. Default value is model name and the entry point.

More typically used keys might be as following (but not limited to and not included by default):

build.commit — commit hash of the build.

build.date — build date of the build.

`scheduler`

Type: list

Scheduler is a component of Sysinspect Master, which is responsible for scheduling the repetitive tasks to call the minions. The aggregate “scheduler” takes a list of tasks. Each task is a list of key/value pairs:

name — name of the task. Type: string. This is a human-readable name of the task. It is used for logging purposes and should be unique.

query — query to be executed on the minion. Type: string. Query is written in a semicolon-separated format
sending the following information: - model name - target scope (e.g. * for all targets)

traits — list of traits to be used for the query. Type: string. E.g. system.os.name:Ubuntu.

interval — interval of the task, i.e. how often the task should be executed. This value can be in seconds, minutes or hours.

interval.unit — unit of the interval. Type: string. This value can be one of the following:

seconds

minutes

hours

days

An example of scheduled tasks:
- name: "Name of your task"

  # Same query as in the command line of SysInspect
  query: "foo/bar;*"
  traits:
    - system.os.name:Ubuntu
    - system.os.version:20.04
  interval: 3
  interval.unit: seconds

- name: "Name of your another task"
  query: "some/model/etc;*"
  interval: 1
  interval.unit: minutes

Example configuration for the Sysinspect Master:

config:
    master:
        socket: /var/run/sysinspect-master.socket
        console.bind.ip: 127.0.0.1
        console.bind.port: 4203
        bind.ip: 0.0.0.0
        bind.port: 4200

        fileserver.bind.ip: 0.0.0.0
        fileserver.bind.port: 4201

        fileserver.models.root: /models
        fileserver.models:
          - my_model
          - my_other_model

        api.enabled: false
        # To enable the embedded Web API, configure TLS first:
        # api.enabled: true
        # api.tls.enabled: true
        # api.tls.cert-file: etc/web/api.crt
        # api.tls.key-file: etc/web/api.key

`datastore.path`

Type: string

Path to the datastore directory, where all data from minions is stored. This is a directory, which is used to store all artifacts from/for minions, for data exchange. This can be anything: a package, a binary, a text file etc.

Default value is /var/lib/sysinspect/datastore.

`datastore.max-size`

Type: string

Maximum size of the datastore directory. This is a string in format <SIZE><UNIT>, where SIZE is a number and UNIT is one of the following:

B (bytes)

KB (kilobytes)

MB (megabytes)

GB (gigabytes)

TB (terabytes)

Default value is 10GB.

`datastore.max-age`

Type: string

Maximum age of the data in the datastore directory. This is a string in format <AGE><UNIT>, where AGE is a number and UNIT is one of the following:

s (seconds)

m (minutes)

h (hours)

d (days)

Default value is 30d.

`datastore.item-max-size`

Type: string

Maximum size of a single item in the datastore directory. This is a string in format <SIZE><UNIT>, where SIZE is a number and UNIT is one of the following:

B (bytes)

KB (kilobytes)

MB (megabytes)

GB (gigabytes)

TB (terabytes)

Default value is 100MB.

Minion

Sysinspect Minion configuration is located under earlier mentioned minion section, and contains the following directives:

`path.root`

Type: string

Typically, Minion if running standard, the root of all data kept by a Minion is defaulted to /etc/sysinspect, same as Master. However, in an embedded and custom systems this might not be possible, especially if the system is usually read-only and writable directories are limited to only a few. In this case root must be set according to the system setup.

`path.id`

Type: string

By default, the minion Id is the /etc/machine-id. However, this file is usually present on a regular Linux server and desktop distributions, but practically never on the embedded systems. For this reason, the alternative location of the machine-id needs to be specified. On many embedded Linux systems and Android, usually /etc is read-only, and very few places are allowed to be written.

This option takes one of the following:

An absolute path to an existing machine-id file

relative keyword, so it is $MINION_ROOT/machine-id, which is /etc/sysinspect/machine-id by default.
id.path: </absolute/path>|relative

`path.sharelib`

Type: string

The location of sharelib directory, which is by default is at the location /usr/share/sysinspect. On most embedded systems those root filesystem is usually read-only, this location can be changed. This directory contains lib and modules subdirectories.

`path.tmp`

Type: string

The location of temporary directory, which is by default is at the location /var/tmp/sysinspect. On most embedded systems those root filesystem is usually read-only, this location can be changed. This directory is used to store temporary files, such as downloaded modules, temporary data etc.

`master.ip`

Corresponds to bind.ip of Master node and should be identical.

`master.port`

Type: integer

Corresponds to bind.ip.port of Master node and should be identical. By default it is set to 4200.

`master.fileserver.port`

Type: integer

Port of Master’s fileserver. By default it is set to 4201.

`offline`

Type: string

Controls whether the Minion continues local execution when the Master becomes unreachable. This is the highest-level policy for network-degraded behaviour.

Accepted values:

independent — the Minion keeps running local work even when the master is gone. Result traffic (events, model results and acknowledgements) is journaled to disk and flushed later when the link returns. Transport recovery happens in the background without destroying the local execution runtime. This is the default.

follow — execution stays coupled to master visibility. Loss of the master may pause or stop normal work while the Minion focuses on reconnection. Intended for tightly controlled environments that require backwards-compatible behaviour.

Summary of the two modes:

Behaviour

follow

independent

Execution

Coupled to master visibility

Continues while master is absent

Delivery

Best-effort with reconnect

Durable journal + background flush

Transport loss

Triggers full instance restart

Degrades transport independently

Use case

Tightly controlled environments

Hostile or unreliable networks

When offline is independent, the durable result journal (described below) becomes a first-class mechanism rather than just a post-restart recovery tool. The Minion will:

journal every result message before attempting network delivery

continue local work even when writes to the master fail

replay all pending results once the transport recovers

keep its module and sensor runners alive through master absence

Warning

independent mode allows the journal backlog to grow while the master is unreachable. Monitor journal.size and the journal eviction warnings to avoid data loss during prolonged outages.

Reliability guarantees by mode:

independent - local execution is allowed to continue while transport is unavailable - durable result traffic is retained and replayed later - delivery is at-least-once, so duplicate result messages are possible - completed cycles are cleared only after the Master returns CycleAck - if reconnect is disabled or reconnect attempts are exhausted, execution

may continue with transport offline and backlog accumulating locally

follow - transport health is treated as execution health - transport/protocol failure may stop the current minion instance - no execution-independence guarantee is made while the Master is absent

For ordinary model commands, the Master now also keeps a durable outbound backlog per target minion. This is separate from the Minion’s result journal.

master -> minion durability contract for ordinary model commands:

submission is durable before live send

delivery is at-least-once

replay occurs after the Master re-establishes the logical session with the Minion and receives fresh Traits

replay identity is per target minion and cycle: mcmd|<minion>|<cycle>

duplicate command delivery is possible during replay, but duplicate local execution is suppressed on the Minion

the Master clears one durable outbound command only after the Minion sends ModelAck for that cycle

master -> minion operational notes:

backlog accounting is per targeted minion, not per original broadcast command

if only some Minions reconnect, only their share of the outbound backlog can be drained

internal cmd://cluster/* commands are not yet covered by this same durable replay contract

Default is independent.

`master.reconnect`

Type: boolean

Sets reconnection to the master (or not). This is a boolean value, which is set to true by default.

`master.reconnect.freq`

Type: integer

Sets the frequency of reconnection to the master. This is a number of times, which is set to 0 by default. There are two options:

0 — infinite reconnection attempts

n — number of reconnection attempts. If the number is reached, the minion will stop trying to reconnect.

`master.reconnect.interval`

Type: string

Interval (seconds) between reconnection attempts. This is a number of seconds, which is set to 5-30 range by default. Possible values are (seconds, between the reconnection attemps):

n — specific amount of seconds

n-n1 — a range of randomly selected seconds within that range

`modules.autosync`

Type: string

Modules are always automatically synchronised at Minion boot. However, it requires full recalculation of each module’s SHA256 checksum and it might take a while, if you have a lot of modules and they are big. Think of this as the startup safety check for modules:

it makes sure the minion has the modules it needs

it optionally verifies them by hashing (SHA256)

More checking = more boot time. Less checking = faster boot, but you trust that files are unchanged.

This value has the following options:

full — safest. Re-hash every module on every boot. Slowest, but detects any change.

fast — balanced. Use cached hashes when available; hash only what is missing. Good for most setups.

shallow — fastest. Only checks that module files exist (no hashing). Best for read-only/embedded boxes. Downside: it will not detect tampering or unexpected edits.

Default is full.

Rule of thumb:

Shared server / security-sensitive: full

Regular servers with many modules: fast

Read-only image / tiny devices: shallow

`modules.autosync.startup`

Type: boolean

Check module checksum on startup. It has two values: - true: check modules on startup - false: do not check modules on startup

Default: true

Warning

Disable this option only if you really know what you are doing. If you disable it, the minion will not check modules on startup, which might lead to unexpected behaviour if modules are changed or tampered with.

`performance`

Type: string

Selects the minion runtime thread profile. This affects Tokio worker threads and blocking threads used by sysminion.

Available values are:

embedded — smallest thread footprint, intended for constrained devices

default — balanced settings for ordinary hosts

server — larger thread pools for throughput-biased deployments

The current thread profiles are:

embedded — register: 1/1, daemon: 2/2

default — register: 2/2, daemon: 4/4

server — register: 4/4, daemon: 8/8

The format above is worker_threads/max_blocking_threads.

Default is default.

Rule of thumb:

Tiny/embedded targets: embedded

General-purpose VM/server: default

Busy hosts with lots of concurrent work: server

`log.forward`

Type: boolean

Forward logs from actions and modules to the main sysinspect log, landing them in the main log file. If disabled, logs from actions and modules will not be forwarded to the main sysinspect log but are kept within their own context inside the returned data and will travel across the whole network back to the master.

Thus, disabling this feature on a large cluster might inflate your network traffic so much that your network admin will start believe in ghosts and aliens.

Warning

Disable this option only if you really know what you are doing.

Default is true

`journal.path`

Type: string

Path to the persistent outgoing journal directory. The journal buffers model/event results on disk before transmission to the master so that no data is lost when the network connection drops. Each record stays in the journal until the master acknowledges the cycle it belongs to.

The journal is part of the Minion’s durable delivery path:

action callbacks (Event) are written before network send

final model callbacks (ModelEvent) are written before network send

model completion notices (ModelAck) are written before network send

on reconnect, any still-pending records are replayed to the Master

a cycle is freed only after the Master sends CycleAck

This means the delivery model is intentionally at-least-once. Replay can cause the same payload to be sent again after reconnect or restart, and the Master is expected to handle that safely.

Default: <managed_db_dir>/journal, which resolves to /tmp/journal on a system-layout install and <root>/tmp/journal on a custom-layout install.

`journal.size`

Type: string / integer

Maximum total size of the outgoing journal in bytes. Accepts plain integers or human-readable strings with an optional unit suffix (B, KB, MB, GB, TB). When the un-acked data exceeds this budget the oldest cycle is evicted automatically. Set to 0 to disable the budget entirely (unlimited).

Warning

Journal eviction is a durability boundary. If the Minion stays offline long enough to exceed this budget, the oldest un-acknowledged cycle is dropped to protect local storage. Sysinspect logs a warning when this happens.

In independent mode, Sysinspect also emits periodic backlog warnings as the journal approaches this limit. The current built-in thresholds are 75%, 90%, and 100% of the configured budget.

Default: 64 MiB.

`offline.backlog.policy`

Type: string

Controls what Sysinspect does when the outgoing journal budget is exceeded during prolonged offline operation.

Accepted values:

evict - drop the oldest un-acknowledged cycle to protect local disk space. This is the default and matches the current implementation.

This policy is applied when journal.size is non-zero and the durable journal grows beyond that byte budget.

Default is evict.

Journaled delivery model

The outgoing journal is not just a cache directory. It is the Minion’s durable network-delivery ledger.

Result traffic that is persisted and replayed:

Event
ModelEvent
ModelAck

Session/control traffic that is not persisted in the journal:

Ehlo
Traits
Ping / Pong
sensor sync requests/responses
Bye and other session-management messages

The design intentionally separates local execution from result delivery:

local work may finish even if the Master is temporarily unavailable
result messages stay in the journal until the Master confirms the cycle
reconnect or restart replays any still-pending result traffic

When a cycle completes locally, the Minion also records a durable local completion marker. This marker exists so that after a hard restart, the Minion can distinguish:

work that never finished locally and must be executed again
work that already finished locally and only needs result replay/delivery

As a result, hard restart recovery follows these rules:

if a cycle had not yet reached durable local completion, rerun is expected
if a cycle had already reached durable local completion, local re-execution is skipped and only journal replay is used to finish delivery

This model is designed for unreliable networks where the Master may disappear temporarily, but the Minion must still converge and later flush its results.

While transport is degraded, Sysinspect also tracks local scheduler backlog. The current DPQ warning thresholds are:

any non-empty backlog
25 or more queued/inflight jobs
100 or more queued/inflight jobs

Example configuration for the Sysinspect Minion:

config:
    minion:
        # Root directory where minion keeps all data.
        # Default: /etc/sysinspect — same as for master
        root: /etc/sysinspect
        master.ip: 192.168.2.31
        master.port: 4200

        # Network-degraded behaviour:
        # offline: independent|follow (default: independent)

        # Outgoing journal (defaults shown):
        # journal.path: /tmp/journal
        # journal.size: 64 MiB

Layout of `/etc/sysinspect`

Ideally, both Master and Minion have the same location of configuration and data collection, which is defaulted to /etc/sysinspect. This directory has many objects stored and has a specific structure and purpose. For more making paths more short, this directory will be referred as $SR (Sysinspect Root).

Common

There are directories that are same on both Master and Minion:

`$SR/functions`

Type: string

Directory, containing custom trait functions. They are meant to be defined on the Master side and then sync’ed to all the minions.

Only on Master

Public and private RSA keys of Master are:

`$SR/master.rsa`

Type: string

Master’s private RSA key.

`$SR/master.rsa.pub`

Type: string

Master’s public RSA key.

`$SR/minion-keys`

Type: string

Public keys from registered minions in format <minion-id>.rsa.pub.

Each registered minion has its own Id. Typically it is /etc/machine-id or automatically generated one, if this file does not exist.

`$SR/minion-registry`

Type: string

A binary cache of minion’s data, such as minion traits, data about currently connected minions etc. This is fully purge-able directory, i.e. data can be freely deleted. However, Sysinspect Master needs to be restarted and all minions needs to reconnect.

Only on Minion

Public and private RSA keys of Master are:

`$SR/master.rsa`

Type: string

Minion’s private RSA key.

`$SR/master.rsa.pub`

Type: string

Minion’s public RSA key.

`$SR/traits`

Type: string

Directory, containing custom static traits of a Minion.

`$SR/models`

Type: string

Directory, containing models.

Configuration

Config Location

Synopsis

Config Section

Common Configuration

log.stream

log.errors

pidfile

Master

socket

console.bind.ip

console.bind.port

console key material

bind.ip

bind.port

cmdb-update

hopstart.*

hopstart.batch

hopstart.network.forward

hopstart.on-start

fileserver.bind.ip

fileserver.bind.port

fileserver.models.root

fileserver.models

fileserver.sensors.root

fileserver.sensors

api.enabled

api.doc

api.bind.ip

api.bind.port

api.auth

api.devmode

api.tls.enabled

api.tls.cert-file

api.tls.key-file

api.tls.ca-file

api.tls.allow-insecure

telemetry.location

telemetry

scheduler

datastore.path

datastore.max-size

datastore.max-age

datastore.item-max-size

Minion

path.root

path.id

path.sharelib

path.tmp

master.ip

master.port

master.fileserver.port

offline

master.reconnect

master.reconnect.freq

master.reconnect.interval

modules.autosync

modules.autosync.startup

performance

log.forward

journal.path

journal.size

offline.backlog.policy

Journaled delivery model

Layout of /etc/sysinspect

Common

$SR/functions

Only on Master

$SR/master.rsa

$SR/master.rsa.pub

$SR/minion-keys

$SR/minion-registry

Only on Minion

$SR/master.rsa

$SR/master.rsa.pub

$SR/traits

$SR/models

`log.stream`

`log.errors`

`pidfile`

`socket`

`console.bind.ip`

`console.bind.port`

`console` key material

`bind.ip`

`bind.port`

`cmdb-update`

`hopstart.*`

`hopstart.batch`

`hopstart.network.forward`

`hopstart.on-start`

`fileserver.bind.ip`

`fileserver.bind.port`

`fileserver.models.root`

`fileserver.models`

`fileserver.sensors.root`

`fileserver.sensors`

`api.enabled`

`api.doc`

`api.bind.ip`

`api.bind.port`

`api.auth`

`api.devmode`

`api.tls.enabled`

`api.tls.cert-file`

`api.tls.key-file`

`api.tls.ca-file`

`api.tls.allow-insecure`

`telemetry.location`

`telemetry`

`scheduler`

`datastore.path`

`datastore.max-size`

`datastore.max-age`

`datastore.item-max-size`

`path.root`

`path.id`

`path.sharelib`

`path.tmp`

`master.ip`

`master.port`

`master.fileserver.port`

`offline`

`master.reconnect`

`master.reconnect.freq`

`master.reconnect.interval`

`modules.autosync`

`modules.autosync.startup`

`performance`

`log.forward`

`journal.path`

`journal.size`

`offline.backlog.policy`

Layout of `/etc/sysinspect`

`$SR/functions`

`$SR/master.rsa`

`$SR/master.rsa.pub`

`$SR/minion-keys`

`$SR/minion-registry`

`$SR/master.rsa`

`$SR/master.rsa.pub`

`$SR/traits`

`$SR/models`