TIR Virtual Machine FAQ

For help with a specific message, see Troubleshooting TIR Virtual Machines.

Basics

How is a TIR Virtual Machine different from a TIR Instance?

A TIR Virtual Machine (VM) is a full operating system with its own Linux kernel and root access, for work that needs operating system changes or kernel modules. A TIR Instance is a container, and of the two, only an Instance takes Spot plans, saves an image, or turns on add-ons at launch. The TIR Virtual Machine overview and the TIR Instance overview describe each product.

Which regions can I use?

You can use Delhi and Chennai. Launching several VMs at once, snapshots, and alerts work in both. To launch a VM, see Create a VM.

What can a VM run on?

A VM runs on a GPU plan, a CPU plan, or your own private cluster. Create a VM explains when to choose each.

Can I use Spot?

No. A VM runs on a GPU plan, a CPU plan, or your own private cluster. For Spot plans, use a TIR Instance.

Is there a web console?

No. You connect over SSH, as Connect to a VM over SSH shows.

What does each VM status mean?

Status	What it means
Waiting	The VM is launching. It turns Running when its operating system is ready.
Running	The VM is up. You can sign in over SSH, and an hourly plan is billed.
Pending	The VM is waiting for compute capacity and starts by itself as soon as capacity frees up.
Stopping	The VM is shutting down after a stop, or being removed after a delete.
Stopped	The VM is off. Its disk and files are kept, and you can start it again.
Error	The launch failed. The VM’s page shows the reason. See Retry a failed VM launch.

The status on a VM’s own page can lag a few seconds behind the VM. To update it, select the refresh button next to it.

How long can a VM stay on Pending?

Up to 24 hours. It starts by itself as soon as capacity frees up, and the wait is free. To give up the wait, use Revoke Request, as in Withdraw a VM that is waiting for capacity.

Can I rename a VM?

Yes. On the VM’s Overview tab, select the pencil next to VM Name, type the new name of up to 50 characters, and confirm it. You need permission to update VMs.

Where do I find a VM’s ID?

In the browser’s address bar on the VM’s details page, after /vm/. Support may ask for it. The support ticket form itself lists VMs by name.

Can I manage VMs through the API?

Yes. You can create, start, stop, and delete VMs through the compute operations in the TIR API.

Creating and launching

What are the rules for a VM name?

A name has up to 50 characters of letters, digits, hyphens, and underscores, starts with a letter, and ends with a letter or digit. The name must differ from every other VM and instance in the project. See Create a VM.

How long does a launch take?

A launch takes about a minute when capacity is free, and you can follow each step on the VM’s page. If a launch gets its machine but stalls before it is ready, TIR stops it as failed after about 15 minutes, and you can then retry it.

Can I launch more than one VM at once?

Yes, up to 10 identical VMs in one request. Their names get a -1, -2 suffix. Set the count in Number of VMs in Create a VM.

What happens if some VMs in a batch fail?

The others carry on. Each VM in the batch succeeds or fails on its own: a VM that fails shows Error in the list, and you retry or delete it separately. TIR checks the credit for the whole batch before it creates any VM. See Retry a failed VM launch.

Why is a plan refused for the image I chose?

The plan must match the image. A CPU plan needs an image with a CPU version, and a GPU plan needs one with a GPU version. A mismatch shows “Invalid SKU ID for the specified Image”. Choose a plan from the other tab, or another image. See Create a VM.

What does Inventory not available mean?

The plan is out of stock right now, so you cannot choose it. A size that is out of stock reads Plan not available. On the create and clone forms, such a plan has a Request button, which sends a request to E2E for that plan. Choose another plan, or check back later. See A plan shows Inventory not available.

Do I need an SSH key to create a VM?

Yes. The form asks for at least one SSH key, and a security group that allows port 22. Both choices are in Create a VM.

Is my account’s verification needed before I launch?

Yes. Until your account has finished its customer validation, Launch VM asks you to complete a billing or country check first. Complete it, then launch again.

Does a clone copy my files?

No. Clone VM builds a new VM from the same settings, such as the disk size, SSH keys, startup scripts, plan, datasets, and shared file systems, with a new disk that has none of the original’s files. To copy the files, create a VM from a snapshot instead. See Clone a VM.

Connecting and SSH keys

Which user do I sign in as?

You sign in as root, with the command in SSH Command under Connection Details on the VM’s Overview tab. See Connect to a VM over SSH.

Can I sign in with a password?

No. A VM accepts SSH keys only. Connect to a VM over SSH has the key command.

Does SSH answer as soon as the VM shows Running?

Running means the operating system is up, and SSH can take a little longer to start answering. If the connection is refused right after the VM turns Running, wait a moment and run the command again.

Can I change SSH keys without restarting the VM?

Yes. The VM keeps running through a key change and picks it up within about 15 seconds. Change keys while the VM is Running or Stopped. See Manage SSH keys on a VM.

Startup scripts

Do startup scripts run only the first time?

No. A startup script runs as root on every boot of the VM, and a change to a script takes effect after the VM restarts. See Add startup scripts to a VM.

How do startup scripts run?

They run one after another, each with bash. They run in the order of their file names inside the VM. Those names come from each script’s ID rather than the name you gave it, so the IDs set the order, whatever you name the scripts. Put steps that must run in order into one script. The scripts after a failed one still run, and a script can still be running after the VM shows Running.

Where can I see what a startup script printed?

Inside the VM. Have the script write its output to a file of your own, and read the file over SSH. The scripts themselves are in /opt/e2e/startup-scripts on the VM. See Add startup scripts to a VM.

How large can startup scripts be?

All of a VM’s startup scripts together can hold about 1 MiB. To set up something larger, have a script download it.

Why can I not delete a startup script?

The script is still attached to a VM. Detach it from every VM first, then delete it. See Add startup scripts to a VM.

Plans and power

Can I change the plan of a running VM?

No. Stop an hourly VM first, then change its plan. The VM then starts again by itself on the new plan, with its disk and files as they were. A committed VM keeps its plan for the whole term. See Change the plan of a VM.

Does a restart run my startup scripts?

Yes. A restart, a start, and a first launch are all boots, and every attached script runs on each one.

Storage

How big can the disk be?

The disk can be from 30 GB to 2,000 GB, or up to your account’s disk limit if E2E has set a lower one. The default is 30 GB. See Resize the disk of a VM.

Can I shrink the disk?

No. A VM’s disk can only grow. To end up with a smaller disk, create a new VM with the size you want. See Create a VM.

Can I make the disk bigger while the VM is running?

Yes. The VM keeps running through a resize. To use the added space, restart the VM or grow the partition and file system yourself. The commands are in Resize the disk of a VM.

Does the disk keep my files when I stop or restart the VM?

Yes. The disk keeps its files across stops, starts, and restarts. See Stop, start, and restart a VM.

What happens to my disk when I delete a VM?

The disk and its contents are deleted with the VM, permanently. Snapshots of the VM stay, and so do the datasets and shared file systems that were mounted on it. See Delete a VM.

Does a VM have a separate workspace disk?

No. A VM has one disk, which holds the operating system and your files. See Resize the disk of a VM.

Can I mount datasets and shared file systems?

Yes. A dataset mounts at /datasets/<dataset name>, and a shared file system at a path you type. Mounting or unmounting restarts a running VM. See Mount datasets and file systems on a VM.

Can a disk dataset be mounted on two VMs?

No. A disk dataset mounts on one VM at a time, read-write. A dataset kept in object storage mounts read-only, and a Shared File System can be mounted by several VMs at once.

Can I mount a Parallel File System?

Yes, on a VM in Delhi. See Mount datasets and file systems on a VM.

Network

Does a VM have a public IP?

Yes. A VM gets a public IP at launch unless you attach a reserved IP. See Attach a reserved IP to a VM.

Does the VM keep its public IP?

Through a restart, yes. After a stop and a start, a floating public IP can change, and deleting the VM releases it. To keep an address, use a reserved IP, or turn the current one into a reserved IP with Convert to reserved. See Attach a reserved IP to a VM.

How many reserved IPs can I have?

You can have five in each region by default, counted across all the projects of your account. Reserving one more shows “You have reached the maximum limit for reserving IPs.” See Delete a reserved IP to free one.

Can I attach a reserved IP when I launch several VMs?

Not at launch. The Reserve IP Address field attaches a reserved IP only when you launch a single VM. For several, attach an address to each VM once it runs. See Attach a reserved IP to a VM.

Which VMs can join a VPC?

VMs on a CPU plan join a VPC directly, which gives them a private address in your network. A VM holds one VPC at a time, and attaching or detaching it restarts the VM. VMs on a GPU plan or a private cluster join through a VPC IP. See Attach a VPC to a VM.

Why does a launch of several VMs fail on my VPC?

The VPC needs a free IP for each VM you launch. Choose a VPC with enough free IPs, or launch fewer VMs.

How many security groups can I attach at once?

You can attach up to five in one go. To attach more, attach the rest in a second go. See Attach security groups to a VM.

Is the default security group open to the internet?

For SSH, yes. It allows SSH on port 22 from every address, and that rule is fixed. To allow only your own addresses, see Limit SSH to your own addresses.

Which ports can I open?

You can open port 80, 443, or any port from 1024 to 65535, for TCP. Ports 22 and 8888 are reserved, so choose another port for a custom rule. See Open a port for an app.

Monitoring

Why does my VM show no memory chart?

A VM launched before E2E added the memory reading can be missing its memory series. The other charts work as usual. See Monitor a VM.

Do I need to install anything to see the charts?

No. TIR reads the CPU, memory, network, and storage figures from outside the VM. On a GPU VM, the GPU figures come from NVIDIA’s DCGM tool. When the image lacks it, the VM downloads it from NVIDIA as it boots, so the VM needs internet access for those figures.

Billing

Do I pay for a stopped VM?

The plan charge stops when you stop the VM, while the disk above the first 30 GB keeps being billed until you delete the VM. See Billing options for TIR Virtual Machines.

How is an hourly VM billed?

An hourly VM is billed by the minute, at the plan’s hourly price, from when the VM becomes ready until you stop or delete it. The time a VM spends waiting or launching is free. For prices, see E2E Networks pricing.

What is a VM charged for, and what is free?

A VM’s own charges are its plan and its disk above 30 GB. SSH keys, security groups, and startup scripts are free. A reserved IP, snapshots, datasets, and shared file systems are billed on their own, and keep billing until you delete them. See Billing options for TIR Virtual Machines.

Is the disk billed?

Yes, for each GB above the first 30 GB, by the hour, from when the VM first becomes ready until you delete it, including while it is stopped. After a resize, the larger size is billed from that moment. See Billing options for TIR Virtual Machines.

How much credit do I need to launch a VM?

On a prepaid account, an hourly launch needs credit for about 24 hours of the plan, and a committed launch needs the full price of the term. A VM on a private cluster skips the check. See Insufficient Credits! when you launch a VM.

Can I stop a committed VM?

No. A committed VM runs for its whole term, and you can restart it. See Stop, start, and restart a VM.

Can I delete a committed VM before its term ends?

Yes, but the days left in the term are not refunded. See Delete a VM.

When is a committed VM charged?

A committed VM is charged once, for the whole term, when it first becomes ready, or when you convert an hourly VM. Which terms you can choose, from one month to three years, depends on the plan. See Billing options for TIR Virtual Machines.

What happens when a committed term ends?

The renewal choice you made takes effect: renew the term, move to hourly billing, or delete the VM. TIR emails the VM’s creator and the account’s primary contact before the term ends, and you can change the choice during the term, as Change the plan of a VM describes.

Can I convert an hourly VM to a committed plan?

Yes, with Convert to Committed. The VM keeps its plan, and it keeps running as it converts. A VM on a private cluster is already covered by the cluster’s plan, so it has no Convert to Committed. See Convert a VM to a committed plan.

Is a reserved IP billed?

Yes. A reserved IP has a monthly fee of its own. For prices, see E2E Networks pricing.

Snapshots

How do I keep a copy of a VM’s disk?

Take a snapshot of it. See Create a VM snapshot and the TIR Snapshots FAQ.