diff options
Diffstat (limited to 'docs/deployment-guides')
| -rw-r--r-- | docs/deployment-guides/README.md | 25 | ||||
| -rw-r--r-- | docs/deployment-guides/deployment-strategies.md | 239 | ||||
| -rw-r--r-- | docs/deployment-guides/deployment-with-centralization-points.md | 121 | ||||
| -rw-r--r-- | docs/deployment-guides/standalone-deployment.md | 139 |
4 files changed, 524 insertions, 0 deletions
diff --git a/docs/deployment-guides/README.md b/docs/deployment-guides/README.md new file mode 100644 index 000000000..18f578857 --- /dev/null +++ b/docs/deployment-guides/README.md @@ -0,0 +1,25 @@ +# Deployment Guides + +Netdata can be used to monitor all kinds of infrastructure, from stand-alone tiny IoT devices to complex hybrid setups combining on-premise and cloud infrastructure, mixing bare-metal servers, virtual machines and containers. + +There are 3 components to structure your Netdata ecosystem: + +1. **Netdata Agents** + + To monitor the physical or virtual nodes of your infrastructure, including all applications and containers running on them. + + Netdata Agents are Open-Source, licensed under GPL v3+. + +2. **Netdata Parents** + + To create [observability centralization points](https://github.com/netdata/netdata/blob/master/docs/observability-centralization-points/README.md) within your infrastructure, to offload Netdata Agents functions from your production systems, to provide high-availability of your data, increased data retention and isolation of your nodes. + + Netdata Parents are implemented using the Netdata Agent software. Any Netdata Agent can be an Agent for a node and a Parent for other Agents, at the same time. + + It is recommended to set up multiple Netdata Parents. They will all seamlessly be integrated by Netdata Cloud into one monitoring solution. + +3. **Netdata Cloud** + + Our SaaS, combining all your infrastructure, all your Netdata Agents and Parents, into one uniform, distributed, scalable, monitoring database, offering advanced data slicing and dicing capabilities, custom dashboards, advanced troubleshooting tools, user management, centralized management of alerts, and more. + +The Netdata Agent is a highly modular software piece, providing data collection via numerous plugins, an in-house crafted time-series database, a query engine, health monitoring and alerts, machine learning and anomaly detection, metrics exporting to third party systems. diff --git a/docs/deployment-guides/deployment-strategies.md b/docs/deployment-guides/deployment-strategies.md new file mode 100644 index 000000000..3be2edbcf --- /dev/null +++ b/docs/deployment-guides/deployment-strategies.md @@ -0,0 +1,239 @@ +# Deployment strategies + + +## Deployment Options Overview + +This section provides a quick overview of a few common deployment options. The next sections go into configuration examples and further reading. + +### Stand-alone Deployment + +To help our users have a complete experience of Netdata when they install it for the first time, a Netdata Agent with default configuration +is a complete monitoring solution out of the box, having all these features enabled and available. + +The Agent will act as a _stand-alone_ Agent by default, and this is great to start out with for small setups and home labs. By [connecting each Agent to Cloud](https://github.com/netdata/netdata/blob/master/src/claim/README.md), you can see an overview of all your nodes, with aggregated charts and centralized alerting, without setting up a Parent. + + + +### Parent – Child Deployment + +An Agent connected to a Parent is called a _Child_. It will _stream_ metrics to its Parent. The Parent can then take care of storing metrics on behalf of that node (with longer retention), handle metrics queries for showing dashboards, and provide alerting. + +When using Cloud, it is recommended that just the Parent is connected to Cloud. Child Agents can then be configured to have short retention, in RAM instead of on Disk, and have alerting and other features disabled. Because they don't need to connect to Cloud themselves, those children can then be further secured by not allowing outbound traffic. + + + +This setup allows for leaner Child nodes and is good for setups with more than a handful of nodes. Metrics data remains accessible if the Child node is temporarily unavailable or decommissioned, although there is no failover in case the Parent becomes unavailable. + + +### Active–Active Parent Deployment + +For high availability, Parents can be configured to stream data for their children between them, and keep the data sets in sync. Child Agents are configured with the addresses of both Parent Agents, but will only stream to one of them at a time. When that Parent becomes unavailable, it reconnects to another. When the first Parent becomes available again, that Parent will catch up by receiving the backlog from the second. + +With both Parent Agents connected to Cloud, Cloud will route queries to either Parent transparently, depending on their availability. Alerts trigger on either Parent will stream to Cloud, and Cloud will deduplicate and debounce state changes to prevent spurious notifications. + + + + +## Configuration Details + +### Stand-alone Deployment + +The stand-alone setup is configured out of the box with reasonable defaults, but please consult our [configuration documentation](https://github.com/netdata/netdata/blob/master/docs/cloud/cheatsheet.md) for details, including the overview of [common configuration changes](https://github.com/netdata/netdata/blob/master/docs/configure/common-changes.md). + +### Parent – Child Deployment + +For setups involving Child and Parent Agents, the Agents need to be configured for [_streaming_](https://github.com/netdata/netdata/blob/master/src/streaming/README.md), through the configuration file `stream.conf`. This will instruct the Child to stream data to the Parent and the Parent to accept streaming connections for one or more Child Agents. To secure this connection, both need set up a shared API key (to replace the string `API_KEY` in the examples below). Additionally, the Child is configured with one or more addresses of Parent Agents (`PARENT_IP_ADDRESS`). + +An API key is a key created with `uuidgen` and is used for authentication and/or customization in the Parent side. I.e. a Child will stream using the API key, and a Parent is configured to accept connections from Child, but can also apply different options for children by using multiple different API keys. The easiest setup uses just one API key for all Child Agents. + +#### Child config + +As mentioned above, the recommendation is to not claim the Child to Cloud directly during your setup, avoiding establishing an [ACLK](https://github.com/netdata/netdata/blob/master/src/aclk/README.md) connection. + +To reduce the footprint of the Netdata Agent on your production system, some capabilities can be switched OFF on the Child and kept ON on the Parent. In this example, Machine Learning and Alerting are disabled in the Child, so that the Parent can take the load. We also use RAM instead of disk to store metrics with limited retention, covering temporary network issues. + +##### netdata.conf + +On the child node, edit `netdata.conf` by using the edit-config script: `/etc/netdata/edit-config netdata.conf` set the following parameters: + +```yaml +[db] + # https://learn.netdata.cloud/docs/agent/database + # none = no retention, ram = some retention in ram + mode = ram + # The retention in seconds. + # This provides some tolerance to the time the child has to find a parent in + # order to transfer the data. For IoT this can be lowered to 120. + retention = 1200 + # The granularity of metrics, in seconds. + # You may increase this to lower CPU resources. + update every = 1 +[ml] + # Disable Machine Learning + enabled = no +[health] + # Disable Health Checks (Alerting) + enabled = no +[web] + # Disable remote access to the local dashboard + bind to = lo +[plugins] + # Uncomment the following line to disable all external plugins on extreme + # IoT cases by default. + # enable running new plugins = no +``` + +##### stream.conf + +To edit `stream.conf`, again use the edit-config script: `/etc/netdata/edit-config stream.conf`. + +Set the following parameters: + +```yaml +[stream] + # Stream metrics to another Netdata + enabled = yes + # The IP and PORT of the parent + destination = PARENT_IP_ADDRESS:19999 + # The shared API key, generated by uuidgen + api key = API_KEY +``` + +#### Parent config + +For the Parent, besides setting up streaming, the example will also provide an example configuration of multiple [tiers](https://github.com/netdata/netdata/blob/master/src/database/engine/README.md#tiering) of metrics [storage](https://github.com/netdata/netdata/blob/master/docs/store/change-metrics-storage.md), for 10 children, with about 2k metrics each. + +- 1s granularity at tier 0 for 1 week +- 1m granularity at tier 1 for 1 month +- 1h granularity at tier 2 for 1 year + +Requiring: + +- 25GB of disk +- 3.5GB of RAM (2.5GB under pressure) + +##### netdata.conf + +On the Parent, edit `netdata.conf` with `/etc/netdata/edit-config netdata.conf` and set the following parameters: + +```yaml +[db] + mode = dbengine + storage tiers = 3 + # To allow memory pressure to offload index from ram + dbengine page descriptors in file mapped memory = yes + # storage tier 0 + update every = 1 + dbengine multihost disk space MB = 12000 + dbengine page cache size MB = 1400 + # storage tier 1 + dbengine tier 1 page cache size MB = 512 + dbengine tier 1 multihost disk space MB = 4096 + dbengine tier 1 update every iterations = 60 + dbengine tier 1 backfill = new + # storage tier 2 + dbengine tier 2 page cache size MB = 128 + dbengine tier 2 multihost disk space MB = 2048 + dbengine tier 2 update every iterations = 60 + dbengine tier 2 backfill = new +[ml] + # Enabled by default + # enabled = yes +[health] + # Enabled by default + # enabled = yes +[web] + # Enabled by default + # bind to = * +``` + +##### stream.conf + +On the Parent node, edit `stream.conf` with `/etc/netdata/edit-config stream.conf`, and then set the following parameters: + +```yaml +[API_KEY] + # Accept metrics streaming from other Agents with the specified API key + enabled = yes +``` + +### Active–Active Parent Deployment + +In order to setup active–active streaming between Parent 1 and Parent 2, Parent 1 needs to be instructed to stream data to Parent 2 and Parent 2 to stream data to Parent 1. The Child Agents need to be configured with the addresses of both Parent Agents. The Agent will only connect to one Parent at a time, falling back to the next if the previous failed. These examples use the same API key between Parent Agents as for connections from Child Agents. + +On both Netdata Parent and all Child Agents, edit `stream.conf` with `/etc/netdata/edit-config stream.conf`: + +##### stream.conf on Parent 1 + +```yaml +[stream] + # Stream metrics to another Netdata + enabled = yes + # The IP and PORT of Parent 2 + destination = PARENT_2_IP_ADDRESS:19999 + # This is the API key for the outgoing connection to Parent 2 + api key = API_KEY +[API_KEY] + # Accept metrics streams from Parent 2 and Child Agents + enabled = yes +``` + +##### stream.conf on Parent 2 + +```yaml +[stream] + # Stream metrics to another Netdata + enabled = yes + # The IP and PORT of Parent 1 + destination = PARENT_1_IP_ADDRESS:19999 + api key = API_KEY +[API_KEY] + # Accept metrics streams from Parent 1 and Child Agents + enabled = yes +``` + +##### stream.conf on Child Agents + +```yaml +[stream] + # Stream metrics to another Netdata + enabled = yes + # The IP and PORT of the parent + destination = PARENT_1_IP_ADDRESS:19999 PARENT_2_IP_ADDRESS:19999 + # The shared API key, generated by uuidgen + api key = API_KEY +``` + +## Further Reading + +We strongly recommend the following configuration changes for production deployments: + +1. Understand Netdata's [security and privacy design](https://github.com/netdata/netdata/blob/master/docs/security-and-privacy-design/README.md) and + [secure your nodes](https://github.com/netdata/netdata/blob/master/docs/category-overview-pages/secure-nodes.md) + + To safeguard your infrastructure and comply with your organization's security policies. + +2. Set up [streaming and replication](https://github.com/netdata/netdata/blob/master/src/streaming/README.md) to: + + - Offload Netdata Agents running on production systems and free system resources for the production applications running on them. + - Isolate production systems from the rest of the world and improve security. + - Increase data retention. + - Make your data highly available. + +3. [Optimize the Netdata Agents system utilization and performance](https://github.com/netdata/netdata/blob/master/docs/guides/configure/performance.md) + + To save valuable system resources, especially when running on weak IoT devices. + +We also suggest that you: + +1. [Use Netdata Cloud to access the dashboards](https://github.com/netdata/netdata/blob/master/docs/quickstart/infrastructure.md) + + For increased security, user management and access to our latest tools for advanced dashboarding and troubleshooting. + +2. [Change how long Netdata stores metrics](https://github.com/netdata/netdata/blob/master/docs/store/change-metrics-storage.md) + + To control Netdata's memory use, when you have a lot of ephemeral metrics. + +3. [Use host labels](https://github.com/netdata/netdata/blob/master/docs/guides/using-host-labels.md) + + To organize systems, metrics, and alerts. diff --git a/docs/deployment-guides/deployment-with-centralization-points.md b/docs/deployment-guides/deployment-with-centralization-points.md new file mode 100644 index 000000000..b3e2b40dc --- /dev/null +++ b/docs/deployment-guides/deployment-with-centralization-points.md @@ -0,0 +1,121 @@ +# Deployment with Centralization Points + +An observability centralization point can centralize both metrics and logs. The sending systems are called Children, while the receiving systems are called a Parents. + +When metrics and logs are centralized, the Children are never queried for metrics and logs. The Netdata Parents have all the data needed to satisfy queries. + +- **Metrics** are centralized by Netdata, with a feature we call **Streaming**. The Parents listen for incoming connections and permit access only to Children that connect to it with the right API key. Children are configured to push their metrics to the Parents and they initiate the connections to do so. + +- **Logs** are centralized with methodologies provided by `systemd-journald`. This involves installing `systemd-journal-remote` on both the Parent and the Children, and configuring the keys required for this communication. + +| Feature | How it works | +|:---------------------------------------------:|:-------------------------------------------------------------------------------------------------------------:| +| Unified infrastructure dashboards for metrics | Yes, at Netdata Cloud | +| Unified infrastructure dashboards for logs | All logs are accessible via the same dashboard at Netdata Cloud, although they are unified per Netdata Parent | +| Centrally configured alerts | Yes, at Netdata Parents | +| Centrally dispatched alert notifications | Yes, at Netdata Cloud | +| Data are exclusively on-prem | Yes, Netdata Cloud queries Netdata Agents to satisfy dashboard queries. | + +A configuration with 2 observability centralization points, looks like this: + +```mermaid +flowchart LR + WEB[["One unified + dashboard + for all nodes"]] + NC(["<b>Netdata Cloud</b> + decides which agents + need to be queried"]) + SA1["Netdata at AWS + A1"] + SA2["Netdata at AWS + A2"] + SAN["Netdata at AWS + AN"] + PA["<b>Netdata Parent A</b> + at AWS + having all metrics & logs + for all Ax nodes"] + SB1["Netdata On-Prem + B1"] + SB2["Netdata On-Prem + B2"] + SBN["Netdata On-Prem + BN"] + PB["<b>Netdata Parent B</b> + On-Prem + having all metrics & logs + for all Bx nodes"] + WEB -->|query| NC -->|query| PA & PB + PA ---|stream| SA1 & SA2 & SAN + PB ---|stream| SB1 & SB2 & SBN +``` + +Netdata Cloud queries the Netdata Parents to provide aggregated dashboard views. + +For alerts, the dispatch of notifications looks like in the following chart: + +```mermaid +flowchart LR + NC(["<b>Netdata Cloud</b> + applies silencing + & user settings"]) + SA1["Netdata at AWS + A1"] + SA2["Netdata at AWS + A2"] + SAN["Netdata at AWS + AN"] + PA["<b>Netdata Parent A</b> + at AWS + having all metrics & logs + for all Ax nodes"] + SB1["Netdata On-Prem + B1"] + SB2["Netdata On-Prem + B2"] + SBN["Netdata On-Prem + BN"] + PB["<b>Netdata Parent B</b> + On-Prem + having all metrics & logs + for all Bx nodes"] + EMAIL{{"<b>e-mail</b> + notifications"}} + MOBILEAPP{{"<b>Netdata Mobile App</b> + notifications"}} + SLACK{{"<b>Slack</b> + notifications"}} + OTHER{{"Other + notifications"}} + PA & PB -->|alert transitions| NC -->|notification| EMAIL & MOBILEAPP & SLACK & OTHER + SA1 & SA2 & SAN ---|stream| PA + SB1 & SB2 & SBN ---|stream| PB +``` + +### Configuration steps for deploying Netdata with Observability Centralization Points + +For Metrics: + +- Install Netdata agents on all systems and the Netdata Parents. + +- Configure `stream.conf` at the Netdata Parents to enable streaming access with an API key. + +- Configure `stream.conf` at the Netdata Children to enable streaming to the configured Netdata Parents. + +For Logs: + +- Install `systemd-journal-remote` on all systems and the Netdata Parents. + +- Configure `systemd-journal-remote` at the Netdata Parents to enable logs reception. + +- Configure `systemd-journal-upload` at the Netdata Children to enable transmission of their logs to the Netdata Parents. + +Optionally: + +- Disable ML, health checks and dashboard access at Netdata Children to save resources and avoid duplicate notifications. + +When using Netdata Cloud: + +- Optionally: disable dashboard access on all Netdata agents (including Netdata Parents). +- Optionally: disable alert notifications on all Netdata agents (including Netdata Parents). diff --git a/docs/deployment-guides/standalone-deployment.md b/docs/deployment-guides/standalone-deployment.md new file mode 100644 index 000000000..5baef805a --- /dev/null +++ b/docs/deployment-guides/standalone-deployment.md @@ -0,0 +1,139 @@ +# Standalone Deployment + +To help our users have a complete experience of Netdata when they install it for the first time, a Netdata Agent with default configuration is a complete monitoring solution out of the box, having all its features enabled and available. + +So, each Netdata agent acts as a standalone monitoring system by default. + +## Standalone agents, without Netdata Cloud + +| Feature | How it works | +|:---------------------------------------------:|:----------------------------------------------------:| +| Unified infrastructure dashboards for metrics | No, each Netdata agent provides its own dashboard | +| Unified infrastructure dashboards for logs | No, each Netdata agent exposes its own logs | +| Centrally configured alerts | No, each Netdata has its own alerts configuration | +| Centrally dispatched alert notifications | No, each Netdata agent sends notifications by itself | +| Data are exclusively on-prem | Yes | + +When using Standalone Netdata agents, each of them offers an API and a dashboard, at its own unique URL, that looks like `http://agent-ip:19999`. + +So, each of the Netdata agents has to be accessed individually and independently of the others: + +```mermaid +flowchart LR + WEB[["Multiple + Independent + Dashboards"]] + S1["Standalone + Netdata + 1"] + S2["Standalone + Netdata + 2"] + SN["Standalone + Netdata + N"] + WEB -->|URL 1| S1 + WEB -->|URL 2| S2 + WEB -->|URL N| SN +``` + +The same is true for alert notifications. Each of the Netdata agents runs its own alerts and sends notifications by itself, according to its configuration: + +```mermaid +flowchart LR + S1["Standalone + Netdata + 1"] + S2["Standalone + Netdata + 2"] + SN["Standalone + Netdata + N"] + EMAIL{{"<b>e-mail</b> + notifications"}} + SLACK{{"<b>Slack</b> + notifications"}} + OTHER{{"Other + notifications"}} + S1 & S2 & SN .-> SLACK + S1 & S2 & SN ---> EMAIL + S1 & S2 & SN ==> OTHER +``` + +### Configuration steps for standalone Netdata agents without Netdata Cloud + +No special configuration needed. + +- Install Netdata agents on all your systems, then access each of them via its own unique URL, that looks like `http://agent-ip:19999/`. + +## Standalone agents, with Netdata Cloud + +| Feature | How it works | +|:---------------------------------------------:|:-----------------------------------------------------------------------------------------------------------------------------------------------------------------------:| +| Unified infrastructure dashboards for metrics | Yes, via Netdata Cloud, all charts aggregate metrics from all servers. | +| Unified infrastructure dashboards for logs | All logs are accessible via the same dashboard at Netdata Cloud, although they are not unified (ie. logs from different servers are not multiplexed into a single view) | +| Centrally configured alerts | No, each Netdata has its own alerts configuration | +| Centrally dispatched alert notifications | Yes, via Netdata Cloud | +| Data are exclusively on-prem | Yes, Netdata Cloud queries Netdata Agents to satisfy dashboard queries. | + +By [connecting all Netdata agents to Netdata Cloud](https://github.com/netdata/netdata/blob/master/src/claim/README.md), you can have a unified infrastructure view of all your nodes, with aggregated charts, without configuring [observability centralization points](https://github.com/netdata/netdata/blob/master/docs/observability-centralization-points/README.md). + +```mermaid +flowchart LR + WEB[["One unified + dashboard + for all nodes"]] + NC(["<b>Netdata Cloud</b> + decides which agents + need to be queried"]) + S1["Standalone + Netdata + 1"] + S2["Standalone + Netdata + 2"] + SN["Standalone + Netdata + N"] + WEB -->|queries| NC + NC -->|queries| S1 & S2 & SN +``` + +Similarly for alerts, Netdata Cloud receives all alert transitions from all agents, decides which notifications should be sent and how, applies silencing rules, maintenance windows and based on each Netdata Cloud space and user settings, dispatches notifications: + +```mermaid +flowchart LR + EMAIL{{"<b>e-mail</b> + notifications"}} + MOBILEAPP{{"<b>Netdata Mobile App</b> + notifications"}} + SLACK{{"<b>Slack</b> + notifications"}} + OTHER{{"Other + notifications"}} + NC(["<b>Netdata Cloud</b> + applies silencing + & user settings"]) + S1["Standalone + Netdata + 1"] + S2["Standalone + Netdata + 2"] + SN["Standalone + Netdata + N"] + NC -->|notification| EMAIL & MOBILEAPP & SLACK & OTHER + S1 & S2 & SN -->|alert transition| NC +``` + +> Note that alerts are still triggered by Netdata agents. Netdata Cloud takes care of the notifications only. + +### Configuration steps for standalone Netdata agents with Netdata Cloud + +- Install Netdata agents using the commands given by Netdata Cloud, so that they will be automatically added to your Netdata Cloud space. Otherwise, install Netdata agents and then claim them via the command line or their dashboard. + +- Optionally: disable their direct dashboard access to secure them. + +- Optionally: disable their alert notifications to avoid receiving email notifications directly from them (email notifications are automatically enabled when a working MTA is found on the systems Netdata agents are installed). |