---
url: https://cubesandbox.com/changelog/v0.7.3.md
---

## 2026.10.12 Release v0.7.3

32 commits from 15 contributors.

### ⭐ Highlights

* **Fixed intermittent network failures when accessing sandboxes**, improving network stability.
* **Improved the S3lvol cache**, making cross-node restore faster.
* **S3lvol hot upgrade: faster I/O recovery during component upgrades**.

### 🎯 Major Features

#### MCP support (E2B compatible)

Sandboxes can now be created with an E2B-compatible `mcp` parameter, which starts an MCP (Model Context Protocol) gateway inside the sandbox:

* The Python / Node / Go SDKs all support it and return the MCP URL and token.
* The gateway can be replaced with any custom implementation that follows the contract. Cube provides a reference implementation under `examples/mcp-gateway`.

### ✨ Enhancements

#### Templates & Storage

* S3lvol hot upgrade: the sandbox I/O pause during an upgrade drops sharply, from about 40 s to about 8 s (#1822).
* Cross-node restore no longer copies the whole disk to the node. A restored sandbox lazy-loads on demand, and repeated data can be accelerated by the local cache (#1898).
* Template creation now checks node validity up front and fails immediately when a problem is found, improving the create experience (#1829).

#### Lifecycle & Control Plane

* Pause is faster: the pause implementation was optimized. A single pause is ~41% faster (150 ms → 88 ms for a 2 GiB sandbox), and ~20% faster under concurrency 50 (#1731).

#### Deployment & Runtime

* PVM host / guest kernel updated from 6.6.69 to 6.6.117 (#1852).

#### Web UI / CubeOps / Other

* Template cards in the Web UI now show the template alias (#1876).

### 🐛 Bug Fixes

#### Lifecycle & Sandbox

* Fixed paused sandboxes losing metadata (labels, template, specs, volume mounts) in Info queries, introduced in v0.7.0 (#1861).
* Fixed occasional clones failing with `no more resource` (#1909).
* Fixed the template list hanging once the cluster had many templates (#1899); the list API and `cubemastercli tpl ls` now support paging (`limit` / `offset`).
* Fixed occasional resource-reclaim issues when deleting a sandbox (#1810).
* Fixed rare block I/O truncation or hangs (#1825).
* Fixed create-from-snapshot rejecting re-sent volumeMounts with "must have exactly one declaration" (#1894).

#### Networking & Security

* Fixed 5-tuple conflicts when accessing a sandbox, which caused intermittent network failures (#1915).
* Improved session reclamation so cleanup is more efficient (#1914).
* Fixed connections being reset when a sandbox NIC was reused (#1919).
* Fixed session state not being refreshed (#1918).

#### Deployment & Upgrade

* Reduced noisy cgroup v2 error logs (#1869).
* Fixed `cubemastercli` accumulating zombie processes; a one-time pod restart clears existing ones (#1860).

#### Other

* Fixed creating a sandbox from a deleted snapshot returning HTTP 500, which SDKs kept retrying (#1868).
* Fixed the Go SDK `Watcher` closing twice under concurrency (#1405).
