FAQ

Frequently asked questions about AnyLog architecture, deployment, networking, data management, querying, and blockchain — organized by topic, with links to full documentation where it already exists.

General

Q: What’s the difference between AnyLog and EdgeLake?

EdgeLake is the open-source version. AnyLog Enterprise is the commercial edition, adding: full blockchain abstraction, aggregations, the security protocol, HA, test suites, training, and technical support.

Feature EdgeLake (Community) AnyLog Enterprise (Subscription)
License Open-source (Linux Foundation) Commercial (contact for pricing)
Virtual Edge Layer ✅ ✅
Rule Engine ✅ ✅
Policy-Based Data Management ✅ ✅
Node Management ✅ ✅
Unified APIs, CLIs, Admin UI ✅ ✅
Supported IoT Connectors ✅ ✅
Blockchain Abstraction Optional add-on ✅
Aggregations ❌ ✅
Security Protocol ❌ ✅
High Availability (HA) ❌ ✅
Test Suites ❌ ✅
Training ❌ ✅
Technical Support ❌ ✅
Publisher node ❌ Not supported ✅

For pricing: AnyLog Pricing. Upgrading is as simple as changing the docker-compose image from anylogco/edgelake to anylogco/anylog-network. (Detailed upgrade steps: not yet documented.)

Q: What are the hardware/resource requirements?

Category Requirement
Operating System Linux (Ubuntu, RedHat, Alpine, SUSE), Windows, Mac
Memory footprint 100 MB available without Docker; 300 MB available with Docker
Databases PostgreSQL (optional); SQLite (default, no install needed); MongoDB (only if blob storage is needed)
CPU Intel, ARM, and AMD supported; single-CPU machines up to large multi-core servers (gateways, Raspberry Pi, etc.)
Network TCP-based network required; overlay networks recommended (Nebula is default); static IP and 3 open ports per node
Deployment options Executable (background process), Docker, or Kubernetes

See Prerequisites for the full compatibility matrix.

Q: What are the node types, and what’s a Cluster?

All AnyLog containers run the same image — configuration determines behavior:

  • Metadata Manager (Master) — emulates a blockchain, coordinating metadata for the network.

  • Operator — stores data collected from devices. Belongs to exactly one Cluster — a policy grouping Operators together for HA and telling the network where data lives.

  • Query — processes/executes queries; any node with the system_query logical database can serve this role.

  • Publisher — routes data to Operators; doesn’t store data itself. Not supported in EdgeLake, and cannot run on the same node as Operator.

  • Generic - this is not documented, but when specifying NODE_TYPE=generic the system will automatically deploy an empty “stand box” with network (TCP, REST and Message broker) configured.

See Nodes for the full breakdown, including cluster policy structure and main/backup roles.

Q: How do I deploy, and do you support zero-touch?

Deployment is entirely configuration based with default values pre-set in the deployment-scripts.

From a Docker or Kubernetes point of view (as an example) the process is:

  1. User defines configurations in a dotenv file
  2. Anylog agent is deployed as a container, with environment variables from the dotenv file
    1. image is pulled from docker
    2. deployment-scripts is cloned (if not set to main)
    3. AnyLog agent is brought up
      1. Environment variables become AnyLog variables ($NODE_TYPE → !node_type)
      2. databases get connected, services start and policies are defined on the blockchain based on the configurations

Zero-touch, with a caveat: AnyLog can’t guess which node type it should be, or which blockchain to connect to — but everything else has default values. In practice this means only 3 parameters are required:

docker run -it -d --detach-keys=ctrl-d --network host \
    -e NODE_TYPE=operator \
    -e LEDGER_CONN=192.167.78.32:32048 \
    -e LICENSE_KEY=XXX

See Getting Started and Installation & Deployment for the full walkthrough.

Q: What is the difference between an Operator and a Publisher?

An Operator stores data locally and responds to queries. A Publisher receives data files and routes them to Operators — it doesn’t store data itself. They cannot run on the same docker / Kubernetes container due to the shared watch directory.

Q: Can I run multiple AnyLog nodes on the same machine?

Yes. Each node needs its own set of ports (TCP, REST, broker) and its own root directory — use different ANYLOG_SERVER_PORT, ANYLOG_REST_PORT, and ANYLOG_BROKER_PORT values per node.

Q: How do I exit a node, or run it in the background?

exit node

For background: Docker with -d (detached), or a systemd service. The local CLI is disabled in background mode — use the Remote CLI or REST API instead.

Network

Q: How do I connect to TCP / REST / the message broker, and what’s the difference?

TCP is peer-to-peer (node-to-node); REST is for external apps/scripts; the broker is MQTT-style pub/sub. See Background Processes → Network services for setup commands and config options for all three.

Q: How do I join a network — is that blockchain sync, or defining a policy?

Different things, often both needed.

  1. Sync pulls in metadata that’s already on the network (you’re catching up to what exists).
  2. Defining a policy publishes new metadata about you (a node/operator/cluster policy) that other nodes then sync in.

See Blockchain Connectivity for sync, and Blockchain Policy for defining one.

Q: My node can’t connect to the network at all — where do I start?

There are 2 questions here: 1. are you connected to TCP service? If no, please review network connectivity 2. do you have blockchain sync configured? If no, please review blockchain connect & sync section

See Troubleshooting → test node — run that first; if it fails, the problem is local to the node, not a network path issue.

Q: My node can’t communicate with other nodes.

  1. When doing get connections do you see network configured
  2. When running test node does the node reply to itself
  3. are firewalls open?

See Troubleshooting → test network, including the common-causes list for DNS/firewall/port-forwarding issues.

Q: What does bind = true vs bind = false mean?

true — the service only accepts connections on the single specified IP.

false — accepts connections on all available IPs on that port. Use false when you want both local and external traffic accepted.

Q: The test network command shows some nodes as unreachable — is that a problem?

Not necessarily — they may be offline, on a different segment, or behind a firewall. Use subset = true in run client to tolerate partial responses:

run client (blockchain get operator bring.ip_port, subset = true) get status

Q: I deployed on AWS (or GCP/Azure) and nodes can’t connect to each other.

Ensure the security group allows inbound TCP on the AnyLog TCP/REST ports from the relevant ranges; on AWS the instance’s public IP is external_ip, private IP is internal_ip; use bind = false; check VPC/subnet routing if using private networking.

Q: What MTU size should I use?

Default 1500 bytes can cause fragmentation in some cloud/overlay setups. If you see dropped connections, try ip link set eth0 mtu 1400 — check your cloud provider’s recommended MTU for your network type.

Q: Nginx (or another Docker service) can’t reach an AnyLog node on the same Windows machine.

Docker on Windows can’t reach the host via localhost — use host.docker.internal and add the alias to your docker-compose.yaml:

extra_hosts:
  - "host.docker.internal:host-gateway"

Data Management

Q: Which databases do you support, and when do I use each?

AnyLog stores sensor /device data in a SQL database and files in a blob or NoSQL type of database. Supported Databases:

  • PostresSQL
  • SQLite
  • MongoDB
  • S3 / S3-like buckets (Minio and Akave)

In addition, AnyLog can coonect to Milvus for vector/similarity search.

See Databases for the full breakdown and how SQL and blob storage relate to each other.

Q: How do I connect to them?

CConnectivity logic / command differs by the database type

Q: How is data partitioned, and why do some tables have a par_ prefix?

Partitioning is a configurable parameter that maybe used to split data into smaller logical tables (by timestamp data type) for better data management and query performance.

See SQL Storage → Table Partitioning for interval syntax, viewing partitions, and dropping them.

Q: Why did drop partition remove more data than I expected?

If you don’t specify how many partitions to keep (keep=N), the system keeps only the newest one — everything else is dropped. If you don’t specify a table name (or use table=*), the drop applies across every table in the database. Always double-check table= and keep= before running this.

Q: Do you support HA / DR?

Yes — at least 2 operators residing in the same cluster gets you both data resilience (replication) and high availability (query failover). See High Availability & Data Resilience for the full mechanism.

Q: Why isn’t my data replicating across the cluster even though distributor/consumer are enabled?

Two likely causes:

  1. A configuration mismatch:
    1. the logical database not being open on the secondary operator
    2. the two operators do not reside against the same relative cluster
  2. Networking / firewall connectivity issues on one of the two nodes

Q: Two of my operators ended up with the same Member ID.

Most likely a hardcoded config value overwrote it, or the operator carried over an old policy from a previous blockchain/platform.

Fix: bring the operator down, remove its persisted local blockchain copy, and bring it back up — confirming no duplicate IP+Port already exists on the network ledger first.

Q: I reset/redeployed an operator on the same IP — do I need a new Member ID?

No. Blockchain identity is meant to be persistent — a reset node should pick up its original Member ID again, not get issued a new one.

Data ingestion

Q: Data isn’t appearing in the database — what do I check?

  1. Streamer and Operator both running? get processes
  2. Watch directory populated? get !watch_dir — check the error directory too: get !err_dir
  3. Database connected? get databases
  4. create_table = true set in run operator, or table created manually?

Q: Why are rows not appearing immediately?

The Streamer buffers by default (60s / 10,000 bytes). Check thresholds with get streaming, or disable buffering: set buffer threshold where write_immediate = true.

Q: What’s the difference between inline column mapping and a mapping policy?

Inline — written directly into run msg client, simple, not reusable. Mapping policy — stored on the blockchain, referenced by ID, reusable across nodes and required for the Operator’s policy parameter.

Q: Can I use dynamic=true together with explicit column mapping?

Yes — dynamic is an alternative to a fixed table_name, not to mapping. Combine dynamic=true with column.* mapping whenever your payload has multiple fields to type/extract.

Q: My dynamic ingestion is putting different sensor readings into one table instead of separate ones.

dynamic=true with no mapping creates one table per sensor automatically. Once mapping is introduced, you need to also split on a per-row table field, or distinct readings under the same topic merge into one table.

Q: My mapping policy JSON is rejected.

The three most common causes: Null instead of lowercase null; a trailing comma on the last key in an object; a doubled opening quote on a string (easy to miss when copy-pasting). See Mapping Policy for the full schema reference.

Q: My mapping policy is published but the Operator isn’t using it.

  1. run msg client is pointing to the wrong mapping policy
  2. There’s a mismatch between the mapping logic and actual data coming in on the given port
  3. run operator isn’t running so content is coming in but not stored.

Q: Why isn’t my blob file (image/video) showing up when I query the SQL table?

there are a few options here:

  1. If there’s a reference point then the issue has to do with file to / file from that transfers the blob from storage node (ie operator) to the query and/or application.
  2. There’s an issue with volume / data reliance directory for blobs on the query node side.

Please review blob extraction in the Edge Data Manager section.

Query

Q: Can period and increments be used in the same query?

Not at this time — increments needs a continuous range to bucket; period is looking for the last occurrence before a timestamp. See SQL Commands → Time-series optimized queries.

Q: Does AnyLog support SQL-style JOINs across tables?

Yes, via include (not a JOIN keyword) — it pulls in data from additional tables beyond FROM. Pair it with extend=(+table_name) to see which table each result row actually came from:

sql my_data include=(other_table) and extend=(+table_name) "select * from my_table where reading_time >= now() -1d"

See SQL Commands for the full option reference.

Q: My query returns no results even though data exists.

Confirm the Operator has the data (get rows count), blockchain sync is current (get synchronizer), operators are reachable, dbms/table names match exactly, the time filter uses NOW() - N hours (not PostgresSQL INTERVAL syntax), and — for POST — that "destination": "network" is set.

Q: Why did SELECT * (or SELECT COUNT(*)) fail even though I can see the data locally?

The Query node needs a table policy on the blockchain to expand * into real columns before distributing the request. Without it, the query fails not because it can’t find the data, but because it can’t determine how the data is organized.

Q: Can I query a specific node only, or profile slow queries?

run client (10.0.0.78:32048) sql my_data "select * from ping_sensor limit 10"

set query log on
set query log profile 5 seconds     # log queries slower than 5 seconds
get query log

Q: What’s the difference between Milvus’s distance for COSINE vs. L2 metrics?

(Pending — needs an owner to confirm before this is documented.)

Q: What do selection and description do in a blob-aware query?

(Pending — needs an owner to confirm before this is documented.)

Blockchain

Q: What’s the difference between a Metadata Manager (master node) and a real blockchain?

A metadata manager (or master) node is a simple AnyLog agent emulating a blockchain locally. A real blockchain (e.g. Optimism) is an actual distributed, trustless ledger. Either works; a real blockchain isn’t required. See Blockchain Connectivity for the full tradeoff table.

Q: How do I connect to a metadata manager node vs. a real blockchain?

Connecting to a Metadata manager node by simply enabling the blockchain sync logic with a specific TCP IP and port of the metadata manager. While an actual blockchain is a bit more comprehensive as it requires wallets, keys and other components needed to conect to an actual blockcahin.

See Blockchain Connectivity for both paths.

Q: Which blockchains does AnyLog actually support?

Documented examples use Ethereum-compatible chains (Optimism shown specifically). (Confirm whether support extends beyond Ethereum-compatible chains — not established anywhere in this doc set.)

Q: What is a policy, and how do I define one?

A blockchain policy is an object defining node information, data information and / or other information that’s critical for the network and is immutable. AnyLog does not store the actual device / sensor data on the blockchain - just the metadata.

See Policies & Metadata for details.

Q: How do I query the blockchain?

The command blockchain get [* | node type] is the basic command struct used to query the blockchain. In addition, the command allows for where and bring conditions.

See Blockchain Commands → Query the Blockchain.

Q: What does the blockchain actually do for AnyLog?

The blockchain layer is the component that informs an AnyLog agent these are the nodes and data in your network.

See Blockchain for farther details.

Q: blockchain test fails.

The local blockchain file is missing, corrupt, or empty.

  • If you have a blockchain service running 
    run blockchain sync 
blockchain reload metadata
  • If you’d like to copy the blockchain file once 
    blockchain seed from [remote tcp ip]:[remote tcp port]

    blockchain seed works against any node with a blockchain (file) on it. The only word of caution is that if there’s an active blockchain service, the file will ultimately be re-synced with the configured ledger conn.

Q: I added a policy, but it’s not visible on other nodes.

Force an immediate sync: run blockchain sync, then confirm on peers with get metadata version.

Q: What’s the difference between blockchain get and blockchain read?

get returns policies after runtime dynamic updates — the normal case. read returns them exactly as received from the ledger, no post-processing. Using the wrong one can make results look inconsistent when nothing’s actually wrong.

Q: What is the correct command to publish a policy?

blockchain insert — the older blockchain push is deprecated:

blockchain insert where policy = !new_policy and local = true and master = !ledger_conn

REST

Q: GET vs. POST vs. PUT — what’s the difference?

  Where parameters go Requires run msg client? Typical use
GET HTTP headers (-H 'command: ...') No CLI-equivalent commands, scripts, curl
POST JSON body (headers become body keys) Yes, if publishing data (broker=rest + topic mapping) Browsers, restricted-header environments, publishing data with mapping
PUT HTTP headers, raw body No — bypasses topic mapping entirely Publishing data directly to a known dbms/table, no mapping needed

Examples:

# GET
curl -X GET 127.0.0.1:32349 -H "command: get status" -H "AnyLog-Agent: AnyLog/1.23"

# POST
curl -X POST 127.0.0.1:32349 -H "Content-Type: application/json" \
  -d '{"command": "get status", "AnyLog-Agent": "AnyLog/1.23"}'

See Using REST for full publish examples (PUT and POST) and Python code.

Q: Should I use User-Agent or AnyLog-Agent?

Both work identically. Use AnyLog-Agent from a browser — User-Agent is a reserved header fetch() can’t set manually, and using it cross-origin can trigger a CORS preflight AnyLog nodes don’t answer by default.

Q: My browser dashboard gets a CORS error.

Either the node isn’t returning Access-Control-Allow-Origin, or a reserved header triggered a preflight. Use the nginx/Flask proxy for production, or switch to AnyLog-Agent and whitelist it:

Access-Control-Allow-Headers: AnyLog-Agent, Content-Type

Q: My SQL query via POST returns empty results but node status works.

Missing "destination": "network" in the body — without it the query only runs on the query node, which holds no operator data.

Notifications

Q: My Telegram/Pushover notification isn’t sending, and there’s no error.

First check the AnyLog-side cause: telegram_body/pushover_body needs the json prefix (telegram_body = json {...}), or the variable stores literal unresolved text instead of real values — the log shows nothing wrong, but the destination silently rejects the malformed body. If that’s not it: verify config (username/auth key), confirm outbound firewall access, confirm the third-party service isn’t down.

Q: My SMS notification via carrier gateway stopped working.

AnyLog’s SMTP-based SMS logic still works — but most carriers have dropped free mail-to-SMS gateway support on their end. This is a carrier policy change, not an AnyLog issue. Use a dedicated SMS API for anything you depend on.

MCP & AI integration

Q: Claude Desktop shows AnyLog MCP tools but all calls fail.

The endpoint is missing the /mcp/sse suffix — correct URL is http://HOST:PORT/mcp/sse.

Q: MCP tools aren’t visible after editing the config.

Fully quit and reopen Claude Desktop — it doesn’t hot-reload. Verify valid JSON (no trailing commas) and confirm mcp-proxy is installed (which mcp-proxy).

Q: Do I need MCP running at all times for a generated dashboard?

No — the recommended approach runs entirely over REST at runtime; MCP is only used once, at generation time, to discover schema/topology. Only the experimental live-MCP dashboard needs MCP at runtime.

Q: My generated dashboard works from curl but fails in the browser with CORS.

Same fix as the general CORS question above — run the nginx/Flask proxy, or use AnyLog-Agent with the node configured to return the required CORS headers.