Cloud Mac User Guide

Solve usage issues by task, from order details to build logs

You do not need to read this guide from start to finish. Open the section for your current task and verify the term, node, connection details, development environment, and log evidence in order. If the issue remains unresolved, include the full context in a console support ticket.

Rent a Cloud Mac now View the quick start

5 task categories 2 plans available 5 available nodes

RUN SHEET

First-use checklist

Ready

Verify the order Model, term, node, and added storage

ORDER

Confirm access Address, account, local network, and credentials

ACCESS

Prepare the toolchain Xcode, command-line tools, dependencies, and cache

BUILD

Preserve diagnostic evidence Timestamp, exact error text, and redacted logs

TRACE

Dedicated Apple Silicon physical node 365 DAYS

Start with your current task

Tell us where you are stuck

Enter a task, status field, or error symptom to jump straight to the relevant checklist. You can also start with one of five categories without reading the sections in order.

Order and instance fields

Confirm what the order defines before deciding whether the instance is abnormal

The order defines the chip, memory, storage, node, and rental term. Console status fields describe the current delivery or runtime stage; do not confuse the two. Actual availability of combined options is determined by the live console response.

Order verification

Five field groups must match exactly

ORDER PROFILE

Rental term: Billed daily, weekly, monthly, or quarterly. Check the current term against the project duration, and do not combine amounts from different terms in one budget.
Physical node: Five nodes are available: Singapore, Japan (Tokyo), South Korea (Seoul), Hong Kong, and the US East Coast. Prioritize team access paths, code source location, and data flow.
Base configuration: ArmVMS M4 includes M4, 16GB RAM, and a 256GB SSD; ArmVMS M4 Pro includes M4 Pro, 64GB RAM, and a 2TB SSD.
Add-ons: Verify +1TB SSD, +2TB SSD, or Thunderbolt 5 daisy chaining separately, and confirm each add-on appears in the same order term as the base configuration.
Payment: Payments support only USDT-TRC20 and Visa / Mastercard / Amex (via Stripe). All charges are settled in US dollars (USD); available gateways are determined by the backend response.

Common statuses

Processing and Connectable

“Processing” usually means the order is still being delivered; “Connectable” means connection details have been generated. After a status change, refresh the order details and use the latest connection fields shown there.

Troubleshooting principle

A correct order does not mean the environment is ready

After confirming the configuration and node, check the system account, free disk space, Xcode version, and project dependencies. Do not create another order to resolve an environment issue.

Check configuration and pricing

Common console fields and verification actions
Field	What it means	What to check	Record if inconsistent
ORDER ID	Unique identifier linked to the order	Whether the ticket, invoice, and instance reference the same number	Full order number
REGION	Region containing the physical node	Whether it matches the node selected at checkout	Node code and displayed name
TERM	Current rental term	Day, week, month, or quarter and its start and end details	Term name and order screenshot
ACCESS	Whether connection details have been generated	Whether the address, account, and access method are complete	Name of the missing field
STATE	Instance runtime stage	Whether the status changes after refresh and access is restored	Exact status text and timestamp

Development environment guide

Make builds reproducible before optimizing them

When moving to a Cloud Mac, first restore consistency across the toolchain, dependencies, and project paths. Do not copy an entire legacy environment at the outset; rebuild from a checklist to reduce architecture mismatches, stale paths, and cache contamination.

Prepare the Xcode project

Record the required Xcode major version, project or workspace entry point, build Scheme, target platform, and minimum OS version. After opening it for the first time, resolve dependencies before running a complete build.

Confirm that project files and submodules are complete
Check that the Scheme can be used for command-line builds
Verify signing configuration separately from the code version

Command-line tools

First confirm the selected developer directory, then check the compiler, version-control tools, and script interpreter. Automation should output the actual tool paths and versions so interactive sessions and runners do not use different environments.

Record the current Xcode selection
Confirm that shell initialization files do not block on interaction
Use explicit paths or controlled environment variables in scripts

Dependency management

Restore dependencies from lockfiles and software manifests whenever possible. After installation, check the Apple Silicon architecture, binary sources, and command resolution paths; do not assume prebuilt files from an older environment still apply.

Keep dependency lockfiles and installation logs
Check whether scripts contain hard-coded paths from the old machine
Put private dependency credentials in controlled variables

Build cache

Cache only reproducible data. Validate the baseline with a cache-free build, then restore dependency caches and derived data step by step. When compilation differences cannot be explained, clear the current task cache before deleting all project data.

Distinguish dependency caches, build caches, and final artifacts
Include toolchain and dependency versions in cache keys
Check cache size and free disk space regularly

CI/CD integration

Use the runner as an auditable physical node

Automation should show who triggered the task, which code version it used, which credentials it read, what artifacts it produced, and what it cleaned up afterward. Dedicated resources reduce contention but do not replace pipeline access controls.

01

Inject minimal credentials

Split credentials by repository, dependency source, and distribution target, granting only the permissions required for the current task. Pass credentials through controlled variables; never write them to the repository, build scripts, or logs.

CREDENTIALS
02

Pin the execution context

Record the commit, branch, Scheme, target platform, Xcode version, and dependency summary. Use separate working directories for parallel tasks to prevent shared derived data from overwriting one another.

CONTEXT
03

Export traceable artifacts

Archives, test reports, symbol files, and build logs should include a task identifier, and their integrity should be checked before export. Retain enough logs to locate failures, even for failed tasks.

ARTIFACTS
04

Clean up the task environment

Remove temporary credentials, mounted files, temporary workspaces, and obsolete intermediate artifacts. Put caches intended for long-term reuse in the designated directory, separate from source code and sensitive materials.

CLEANUP

AI experiment guide

Plan the model, storage, and session before starting long-running tasks

The main risks in AI experiments are usually not the launch command, but duplicated model files, growing disk usage, interrupted sessions, and results that are not exported promptly. Establish data boundaries first to avoid discovering insufficient space halfway through a task.

Capacity planning

Organize model files into three directory layers

SOURCE Original model

Keep the source, version, license information, and checksum summary. Store one copy of each original file to avoid duplication across experiment directories.

CACHE Rebuildable cache

Place download caches, conversion caches, and temporary shards in a separate directory. When space is tight, they can be removed and regenerated.

OUTPUT Experiment results

Archive checkpoints, metrics, logs, and exported files by task ID, then move them to long-term storage when the task is complete.

Long-running sessions

Run training, conversion, or batch inference outside temporary graphical sessions. After launch, record the process ID, working directory, log path, and recovery method, then test whether the task continues after disconnection.

Keep writing logs to a file
Use explicit checkpoint intervals
Reserve room for output growth

Backup boundaries

Back up code, model provenance, non-rebuildable checkpoints, and final results separately. Download caches, temporary shards, and reproducible intermediate files should not consume the same backup resources.

Get advice on AI workload selection

Troubleshooting decision tree

Check one layer at a time; change only one variable

After each layer, change one condition and record the result. Network, credentials, system, disk, and build logs have a clear order; skipping preliminary checks can make an access issue look like a development-environment problem.

Can the local network reach the connection address?

Check that the address is complete, whether the current network restricts the relevant connection, and whether the symptom is the same after switching networks. Record the client OS, network type, and time of the error.

Failed Keep the address fields and exact network error text

Are the connection credentials from the current order details?

Reopen the order details in the console and verify the account, address, and credential entry method character by character. Do not use old screenshots, browser history, or connection details from another instance.

Failed Record missing fields; do not submit the credential text

Does the instance status allow a session to be established?

Refresh the instance details and check whether the status fields have changed. If the status does not match the actual access result, record the order number, node, exact status text, and time first observed.

Abnormal Submit the status fields and reproduction steps

Is there enough working space on disk?

Check the project directory, derived data, dependency caches, model files, and temporary artifacts. Once a space issue is confirmed, clear rebuildable caches first; do not delete build results that have not been exported.

Insufficient Record total space, available space, and major directories

Can the build log identify the first real error?

Work backward from the task exit status to find the first error; do not capture only the last line. Keep the command, toolchain version, commit, and error context, and remove tokens, private keys, and other sensitive information.

Unresolved Redact the logs and submit a support ticket

Escalate to support

When self-service does not work, submit context that can go straight into troubleshooting

For existing-order issues, link the order through a console support ticket whenever possible. For pre-sales configuration questions or when you cannot access the console, email support@armvms.com. Continue adding materials in the same conversation to avoid creating duplicate requests.

Required information

Order and node details

Provide the full order number, node name, base configuration, and add-ons. If the issue affects a specific instance, include the current status field shown in the console.

Timing and reproduction

Describe the first failing action

State when it happened, the steps taken, the expected result, and the actual result. If you tried multiple times, specify which conditions changed and which remained the same.

Diagnostic attachments

Attach the exact error and redacted logs

Copy the complete error message and include the relevant build steps and context. Screenshots should show field names, but must hide all sensitive credentials and project secrets.

How should I add new logs or reproduction results after submitting?

Reply in the same ticket or email thread, noting when the new test occurred, which condition changed, and the result. This lets the service team compare events chronologically without rebuilding the context.

When should I use a console support ticket first?

Use a console support ticket first for existing-order issues such as order status, billing, nodes, instance access, and delivery fields. Tickets can be linked to the order number, reducing back-and-forth verification.

What information should I prepare for a pre-sales consultation?

Describe the project type, number of concurrent tasks, required Xcode environment, estimated storage, rental term, and target node. For large projects or large-model inference, highlight peak memory and model-file size.

Prepare your troubleshooting materials

Link the order in your ticket so the issue enters the right context

Include the order number, node, time, reproduction steps, and redacted logs. For pre-sales questions, you can also review the two configurations first, then provide the project duration and resource requirements.