Skip to main content
Migration Notice
We're migrating documentation from the old portal into this one. Some things may look a little different or out of place in the meantime — we know, and we're working to get it right. If something's unclear or doesn't look right, let us know.

Agent Dashboard — TCP In-Use (Established)

What this chart shows​

TCP In-Use — the number of TCP sockets in ESTABLISHED or other active states on the agent host, read from /proc/net/sockstat (inuse field) on every heartbeat (~30 s + 0–10 s jitter).

This counts all active TCP connections on the host, including S3 connections, ClickHouse connections, console heartbeat connections, and any other services running on the same machine.

Linux only​

This metric requires /proc/net/sockstat.

How to read it​

Value rangeInterpretation
10–50Idle or light load — console heartbeat + a few S3 keep-alives
50–200Active copy job — S3 connection pool in use
200–500Heavy parallel copy — multiple jobs, large thread counts
> 500Very high — check for connection leaks or unusually large thread/connection pools

Relationship to copy throughput​

S3 TransferManager (TM) opens parallel connections per upload part. With copy_threads_per_job = 16 and maxConcurrency = 64 (threads × 4), a single busy agent can hold 100+ S3 connections open simultaneously. A rising TCP In-Use count that correlates with active jobs is expected and healthy.

Polling interval​

Sampled on every heartbeat (~30 s + 0–10 s jitter) from the Linux kernel's running socket count.

Tips​

  • Combine with TCP TIME_WAIT to distinguish healthy active connections from sockets draining after jobs complete.
  • A TCP In-Use value that stays elevated well after all jobs finish may indicate a connection leak in the S3 client — check the agent log for connection-pool warnings.