---
last_updated: "2026-10-30T12:22:12.000Z"
title: "Query running merges with system.merges"
excerpt: "You can now query system.merges to see the merges running in your Workspace. Also: a clearer Cluster page, MCP errors that clients can retry, and fixes for imports, Tokens and Workspace admins."
date: "2026-10-30"
---

# Query running merges with system.merges

A long merge can hold up a deployment, and until now you couldn't see which one. You can now query `system.merges` from a Workspace to list the merges and mutations running on your Data Sources:

```sql
SELECT table, elapsed, progress, num_parts, is_mutation
FROM system.merges
```

The table shows only your Workspace's database and a limited set of columns, including `table`, `elapsed`, `progress`, `num_parts`, `is_mutation`, `rows_read` and `merge_type`.

`system.columns` also stopped returning native ClickHouse® columns that the system table replacements reject, so tools that build queries from it no longer get a 400.

## Clearer Cluster page and Overview

The replicas table on the Organization Cluster page now has a header for every column, roomier Load, CPU and Memory bars, and long Cluster names no longer run into the next column. On the Overview, Cluster health now shows the same Cluster names as the Cluster page instead of internal IDs.

## Improvements and bug fixes

- [All]: Added a random delay to the `Retry-After` header on rate-limited API requests, so clients that wait and retry no longer hit the limit again at the same second.
- [All]: Changed the Tinybird MCP Server to return tool errors with `isError: true`, so MCP clients and agent frameworks run their error handling and retries. The row limit note now appears only when a result reaches the limit.
- [All]: Fixed queued import jobs using outdated JSONPaths after you redeploy a Data Source, which made S3 backfills fail with `Schema mismatch`. The error now lists the columns that don't match.
- [All]: Changed role updates to reject demoting the only admin of a Workspace, so a Workspace always keeps an admin.
- [All]: Fixed on-demand copy jobs in Developer Organizations timing out after 30 s. They now get the documented 30 minute limit.
- [All]: Fixed the Workspaces list in the Add member dialog not scrolling with the mouse wheel or trackpad.
- [All]: Changed the connectors API to return a 400 instead of a 500 when a request body isn't valid JSON.
- [Classic]: Fixed `tb push` removing the row-level filter from an existing Token's read scopes when the Token was updated by name.
- [CLI]: Improved the `tb migrate-to-forward` error when a live Classic Pipe fails to parse. It now names the Pipe instead of showing only `Empty query`.