Capacity and limits
Every limit the server applies, its default, and what happens when it is reached. Limits you can change are marked with their configuration key or flow field.
Messages
| What | Limit | At the limit |
|---|---|---|
| Message body (API, every source) | 10 MiB | API and http source 413; mllp source AR; file source moves the file to moveTo/rejected (or skips it); database source reports the row as source.database.refused and leaves it unmarked. |
| Messages processed at once | 32 — processing.maxConcurrent (1–10000) |
A new message waits up to processing.waitMs (default 5000, up to 60000), then is refused as busy: API/http 503 + Retry-After: 1, mllp AE, file and database sources try again a moment later. See processing. |
| Delivery attempts per destination | 5 — delivery.maxAttempts (1–1000) |
The message is dead-lettered; requeue it when the destination is back. |
| Retry delay | 1 s — delivery.backoffBaseMs, doubling per attempt, at most 1 minute |
— |
| HTTP destination reply | 1 MiB read | Larger replies still count as delivered; the reply is not stored or used. |
| Destination timeout | 30 s — timeoutMs per http, mllp, or database destination (1000–120000) |
The attempt fails with lastCode net:timeout and is retried. |
Sources
| What | Limit | At the limit |
|---|---|---|
| http source request read | 60 s — readTimeoutMs (1000–600000); headers within 10 s |
The connection is closed. |
| mllp source connection idle | 5 minutes | The connection is closed; the sender reconnects. |
| mllp source frame | 15 minutes to arrive once started; 10 MiB | Larger frames are read to their end and answered AR. |
| mllp TLS handshake | 30 s | The connection is closed. |
| mllp destination ACK | 1 MiB | mllp:ack-too-large; retried. |
| File source files per poll | 100 | The rest are read in back-to-back polls. |
| File source recursion | 32 directory levels | Deeper directories are not read. |
| File source settle time | 1 s unchanged before a file is read | The file is read once it stops changing. |
| File source interval | 1 s — pollIntervalMs (100–3600000), or a cron schedule |
— |
| Database source rows per poll | 100 — maxRows (1–10000), and at most 64 MiB of row data |
The rest are read in back-to-back polls. |
| Database source interval | 5 s — pollIntervalMs (1000–3600000), or a cron schedule |
— |
| Database query or update | 30 s — timeoutMs (1000–120000) |
The poll fails (source.database.failed) and is retried. |
Formats
| What | Limit | At the limit |
|---|---|---|
| XML nesting / elements | 256 levels / 100,000 elements | The message is refused as not well-formed XML (400, AR). |
| Delimited text | 100,000 rows / 1,000,000 values | The message is refused as not valid delimited text. |
| HL7 v2 versions | 2.1–2.9 (with inputFormat: hl7v2) |
The message is refused (400, AR). |
API and CLI
| What | Limit | At the limit |
|---|---|---|
| Message search page | limit 1–1000 (default 100) |
400; use offset for the next page. |
| Message archive export / import | 10,000 messages per export; imports up to 100 MiB (512 MiB uncompressed) | 400 / 413. |
| Flow import, bulk update, config documents | 50 MiB per request or file | 413. |
| Config map, scripts, settings bodies | 10 MiB | 413. |
| User administration bodies | 1 MiB | 413. |
| Lookup keys / values | 512 characters / 64 KiB | 400. |
| Lookup batch | 1–1000 keys | 400. |
| Lookup entries per request | 1000 by default, limit up to 10000 |
400 above 10000. |
| Events per request | 1000 by default | — |
| Statistics series | 1000 points per request by default, limit up to 10000 |
400 above 10000. |
| Message trends | 1000 buckets per request | 400; use a longer interval. |
weavster deadletter list |
1000 messages per call | — |
| CLI shell and batch-script lines | 1 MiB per line | The shell stops with an error. |
Retention
| Data | Where | Kept |
|---|---|---|
| Messages, their content, metadata, and attempt records | The store (store.dialect) |
Until you remove them (DELETE /api/v1/messages/{id}, or many at once). Nothing is pruned unless you configure prune (maxAgeHours, maxMessages), which removes finished messages on a schedule; otherwise size your database for your traffic, or remove old messages yourself, for example with a nightly DELETE /api/v1/messages?to=…. With memory, until the server stops. |
| Flow definitions, users, config items | The store | Until changed or removed. |
| Events | Server memory and the store | The API shows the latest 10,000, reloaded after a restart with PostgreSQL; the store keeps every event until prune.eventMaxAgeDays. |
| Statistics (current, lifetime) | Server memory and the store | Until reset or the flow is deleted; saved with every sample and at stop, so kept across restarts with PostgreSQL. With memory, until the server stops. |
| Statistics over time | Server memory and the store | stats.retentionHours (default 24, up to 8760), at most 100,000 samples per flow; in memory at most 1,000,000 in all, the newest of which are loaded again at start. |
| Audit log | Server stderr and the store | On stderr, as long as your log collection keeps it. In the store, until prune.auditMaxAgeDays (default: kept); with memory or disabled, the newest 100,000 entries until the server stops. See Audit log. |
To estimate the store's size, count each message's body several times: the original, the stored received form, the transformed output, and a reply when a destination returns one.