changing number of CPUs for an instance

[VasiliyS]

Is it possible to change number of CPUs available to an existing instance? What about memory?

[Saviq]

Hi @VasiliyS, I'm afraid it's not, not easily at the moment. #756 designs the experience, but it's not yet implemented.

Depending on platform, you could just edit the VM in the hypervisor (Hyper-V, VirtualBox), or stop Multipass and edit its instance database.

Let us know what your environment is and we can guide you there :)

[VasiliyS]

@Saviq - Thank you. I'm on a Mac. Where can I find the instance database?

[Saviq]

@VasiliyS here are the steps:

# stop multipassd
sudo launchctl unload /Library/LaunchDaemons/com.canonical.multipassd.plist

# edit /var/root/Library/Application Support/multipassd/multipassd-vm-instances.json
# you'll need sudo for that

# start multipassd again
sudo launchctl load /Library/LaunchDaemons/com.canonical.multipassd.plist

The format should be quite easy to grok:

{
    "primary": {
        "deleted": false,
        "disk_space": "5368709120",
        "mac_addr": "52:54:00:5c:f4:14",
        "mem_size": "1073741824",
        "metadata": {
        },
        "mounts": [
        ],
        "num_cores": 1,
        "ssh_username": "ubuntu",
        "state": 0
    }
}

You want to edit num_cores for CPU and mem_size for memory :)

[VasiliyS]

@Saviq - Just tried it. The instance doesn't start - the cli times out ("timed out waiting for response"). multipass info tells me that the state is unknown and the json file has state set to 8. multipass stop ends with "process error occurred Crashed". The state key in the json file is now 0.


After stopping the multipassd, editing the num_cores key value back to 1 and re-starting the daemon, the instance starts fine.

[Saviq]

@VasiliyS hmm can you look in /Library/Logs/Multipass/multipassd.log for any errors? Does a new instance with 2 CPUs (multipass launch --cpus 2) work fine?

[VasiliyS]

@Saviq - I was able to create a new instance with 2 cpus, worked fine.

I'm attaching 2 files with excerpts from the log. One where I've edited the cpu key to 2 and the instance doesn't start.
2_cpus_doesnt_start.log

The second one where I've changed the setting back to 1 cpu and it works.
1_cpu_starts.log

[VasiliyS]

@Saviq - Update!
I've just tried something. The instance actually starts fine, it's the cli that timeouts.
multipass start exits with a timeout message, but I'm still able to ssh into the instance, etc.


The weird thing is , now even the instance that I've launch'd with --cpus 2 option from scratch doesn't start normally (i.e. multipass start timeouts), but the instance actually works fine. That gave me an idea to try again with editing the num_cores key and ignore the timeout message. Now that I know where the id_rsa is, I can login directly even when multipass shell/info/etc don't work normally. Oh well...

[townsend2010]

Hi @VasiliyS,

It looks like you are hitting #1122 when it's using 2 cpu's. The next version of Multipass has a fix for that. We have a release candidate for 0.9 up and can be found at https://github.com/CanonicalLtd/multipass/releases/download/v0.9.0-rc/multipass-0.9.0-rc.425+g37fa4305.mac-Darwin.pkg if you are interested.

[VasiliyS]

Hi @townsend2010 - thanks! The RC works just fine.
I guess now my original request has been met. I'm now able to change the number of CPUs for an existing instance.

Would be great to add a command to update the instance configuration via the cli, though. It's probably a common need, especially if one uses multipass to create a K8S machine and then discovers that the original resources are not sufficient.

[Saviq]

@VasiliyS glad you got it going. Yes, we'll allow changing the number of CPUs in due course :)

[sourcedelica]

And memory please.

[kramik1]

Just to save some people some time, the below is for Ubuntu 19.10.

sudo systemctl stop snap.multipass.multipassd.service
sudo vi /var/snap/multipass/common/data/multipassd/multipassd-vm-instances.json
sudo systemctl start snap.multipass.multipassd.service

[denizzzka]

@kramik1 these settings will be overwritten by previous values if snapcraft clean will be called

[denizzzka]

I drive into same problem with RAM setting: it is just too low for my snap (only 2Gb) and I can't to build anything.

[townsend2010]

Hi @denizzzka,

There are some Snapcraft environment variables you can set to tweak memory, CPU count, and disk size. They are:

SNAPCRAFT_BUILD_ENVIRONMENT_CPU
SNAPCRAFT_BUILD_ENVIRONMENT_MEMORY
SNAPCRAFT_BUILD_ENVIRONMENT_DISK

Hope this helps!

[denizzzka]

Thanks @townsend2010

I tried this variables. Looks like it work inside of multipass VM, but VM itself still uses 2 cores and 2Gb of RAM (I checked this by atop utility)

[townsend2010]

Hey @denizzzka,

Hmm, it's not supposed to work that way. 😕

I'll run some experiments in a while and if it's not working as it should, I'll ping the Snapcraft folks as they are the ones responsible for setting up these Multipass instances.

[denizzzka]

Thanks!

Please note that I did not change /var/snap/multipass/common/data/multipassd/multipassd-vm-instances.json settings file, because this makes no sense - it is automatically overwritten every time if snapcraft repository is cleaned

[townsend2010]

Hey @denizzzka,

It seems to be working as expected here. Here's what I did on my Ubuntu host machine:

$ SNAPCRAFT_BUILD_ENVIRONMENT_CPU=4 SNAPCRAFT_BUILD_ENVIRONMENT_MEMORY=8G snapcraft
Launching a VM.
'SNAPCRAFT_BUILD_ENVIRONMENT_CPU' was set to '4' in the environment. Changing the default allocation upon user request
'SNAPCRAFT_BUILD_ENVIRONMENT_MEMORY' was set to '8G' in the environment. Changing the default allocation upon user request
...

I can clearly see the snapcraft instance used to build my snap is now using 4 cores and has 8GB allocated to it. Did you snapcraft clean and ensure the instance was gone via multipass list before trying the environment variables?

[denizzzka]

Offtopic:
I removed everything from my Debian and start again. After snapcraft was installed I faced with:

$ snapcraft
cannot snap-exec: cannot exec "/snap/snapcraft/4126/usr/bin/python3": no such file or directory

It worked previously. The experiments are postponed indefinitely.

[denizzzka]

Ok, I sucessfully reinstalled it. Works! Thanks @townsend2010 !

[acefei]

is there any effect way on windows?
The following steps (run as Administrator) are what I did, but the modification doesn't work, the VM started as before.

multipass.exe stop --all
Stop-Service -Name Multipass
notepad.exe 
 C:\Windows\System32\config\systemprofile\AppData\Roaming\multipassd\multipassd-vm-instances.json
Start-Service -Name Multipass
multipass.exe start --all

multipass version

multipass  1.1.0+win
multipassd 1.1.0+win

[Saviq]

@acefei on Windows you can go to the Hyper-V manager (or psexec.exe -s -i VirtualBox, if using that) and change it there.

[Saviq]

@acefei you can do that with PowerShell's Set-VMProcessor cmdlet.

[huangxuankai]

I try two solution to chage cpu, mem, disk space, learn from @kramik1
host: ubuntu 16.04
guest: ubuntu 18.04
snap multipass: 1.2.0

sudo systemctl stop snap.multipass.multipassd.service

// here i only change memory and cpu 
sudo vi /var/snap/multipass/common/data/multipassd/multipassd-vm-instances.json
sudo qemu-img resize /var/snap/multipass/common/data/multipassd/vault/instances/instance_xxx/instance_xxx.img +10G

sudo systemctl start snap.multipass.multipassd.service

multipass start instance_xxx
// here runing ok
multpass info instance_xxx

but key disk_space in file multipassd-vm-instances.json remain not change, I don't know if there will be problems in the future

another solution
i stop all instance on multpasss
set sudo multipass set local.driver=libvirt
then use virt-manager to change cpu and mem
disk space change use sudo qemu-img resize /var/snap/multipass/common/data/multipassd/vault/instances/instance_xxx/instance_xxx.img +10G
finally, start instance, and it run well

[santiagomoneta]

Can anyone share some light on how to do this on MacOS (Catalina). Thanks

[Saviq]

Hi @santiagomoneta, one of the first comments (#1158 (comment)) has the steps.

[hemna]

Seems like there should be a much easier way to do this from the multipass command line itself.
Like,
multipass stop
multipass update -c 4
multipass start

[Saviq]

@hemna we have something like this scheduled for the near months. See #756 for details of what we want to go for.

[mconner]

@Saviq

If we modify the cpu or memory in Hyper-V Manager, the values in multipassd-vm-instances.json are not updated. Should we edit it to match?

[Saviq]

@mconner for Hyper-V and VirtualBox there's no need - they hold their own VM configuration.

[Saviq]

FWIW we have scheduled basic instance modification this cycle, we'll be starting on it soon.

[oldthreefeng]

Niubility...

[tuxerrante]

@magmine then you want to follow https://multipass.run/docs/using-virtualbox-in-multipass-windows and modify the instance with either VBoxManage or the VirtualBox GUI. The Set-VMProcessor cmdlet only works for Hyper-V instances.

Note that when you can't keep the GUI open, otherwise Multipass will have trouble managing the instance (VirtualBox only allows a single entity to manage VMs at any time).

After switching to VBox I can't see them anymore, nor in multipass list nor in VBox GUI, even after other commands of that tutorial.
Should I delete them and recreate them with the VBox driver?
Can I choose the image directory such that I can manually import them in VBox?

EDIT:
I was able to do it through the Hyper-v manager

[matejk]

sudo vim /var/root/Library/Application Support/multipassd/multipassd-vm-instances.json

You either need to escape the space (Application\ Support), or put the whole path in quotes.

With multipass 1.8.1 on macOS Monterey the configuration file is

/var/root/Library/Application\ Support/multipassd/qemu/multipassd-vm-instances.json

[sijakret]

For mac OS this is what helped me: https://www.atpeaz.com/change-multipass-instance-memory-and-cpu-allocation-after-launch/

it was important to unload the multipass service first:

sudo launchctl unload /Library/LaunchDaemons/com.canonical.multipassd.plist

then modify multipassd-vm-instances.json

@VasiliyS here are the steps:

# stop multipassd
sudo launchctl unload /Library/LaunchDaemons/com.canonical.multipassd.plist

# edit /var/root/Library/Application Support/multipassd/multipassd-vm-instances.json
# you'll need sudo for that

# start multipassd again
sudo launchctl load /Library/LaunchDaemons/com.canonical.multipassd.plist

The format should be quite easy to grok:

{
    "primary": {
        "deleted": false,
        "disk_space": "5368709120",
        "mac_addr": "52:54:00:5c:f4:14",
        "mem_size": "1073741824",
        "metadata": {
        },
        "mounts": [
        ],
        "num_cores": 1,
        "ssh_username": "ubuntu",
        "state": 0
    }
}

You want to edit num_cores for CPU and mem_size for memory :)

then restart

sudo launchctl load /Library/LaunchDaemons/com.canonical.multipassd.plist

otherwise it would always reset the changes I made. (multipass stop was not enough in my case!)

[jbosse]

I can't find multipassd-vm-instances.json anywhere on my M1 Mac.

[Saviq]

@jbosse it should be /var/root/Library/Application Support/multipassd/qemu/multipassd-vm-instances.json - if it doesn't exist… did you launch any instances?

[jbosse]

@Saviq Thank you. Mac really doesn't want you to see it in the finder or open it in VSCode. I could only make it open in nano. I had given up though and just made a new instance passing the --mem and --cpus options I wanted.

[colemancda]

multipassd-vm-instances.json doesn't exist on my MacBook Pro (M1 Pro) either.

[jbosse]

multipassd-vm-instances.json doesn't exist on my MacBook Pro (M1 Pro) either.

@colemancda Try sudo ls /var/root/Library/Application\ Support/multipassd/qemu/ to see the contents or sudo nano /var/root/Library/Application\ Support/multipassd/qemu/multipassd-vm-instances.json to view the file.

[pauljsymonds]

@VasiliyS here are the steps:

# stop multipassd
sudo launchctl unload /Library/LaunchDaemons/com.canonical.multipassd.plist

# edit /var/root/Library/Application Support/multipassd/multipassd-vm-instances.json
# you'll need sudo for that

# start multipassd again
sudo launchctl load /Library/LaunchDaemons/com.canonical.multipassd.plist

The format should be quite easy to grok:

{
    "primary": {
        "deleted": false,
        "disk_space": "5368709120",
        "mac_addr": "52:54:00:5c:f4:14",
        "mem_size": "1073741824",
        "metadata": {
        },
        "mounts": [
        ],
        "num_cores": 1,
        "ssh_username": "ubuntu",
        "state": 0
    }
}

You want to edit num_cores for CPU and mem_size for memory :)

How would I do this on Ubuntu?

# stop multipassd
sudo launchctl unload /Library/LaunchDaemons/com.canonical.multipassd.plist

# edit /var/root/Library/Application Support/multipassd/multipassd-vm-instances.json
# you'll need sudo for that

# start multipassd again
sudo launchctl load /Library/LaunchDaemons/com.canonical.multipassd.plist

[Saviq]

Anyone stumbling here, you can now try the implementation in action in #2353, with snaps under edge/pr2353 channel and Windows and macOS packages.

[muhammad-asn]

Thank you this solution works on Mac

[lulor]

@Saviq Thank you for the update, it works great!
May I ask why both the number of cores and the memory can only be increased? It makes perfect sense for the disk, but to me it's less obvious for the other two parameters

[Saviq]

@lulor the primary reason are blueprints, which declare the minimum memory and CPU cores meaningful for a particular use case. That said, we're reconsidering whether we should be that limiting, especially that if you make a mistake and give too much, there's currently no way back.

[lulor]

@Saviq I see, thank you for the explanation

[almontes1987]

I got it using another way. If you're using the VirtualBox driver, open the VirtualBox with the root account. There will be a new tab for Multipass with the instance already created. Just turn off the multipass instance, change the setting on the VirtualBox window and then turn on your multipass instance again.

[daviewales]

# stop multipassd
sudo launchctl unload /Library/LaunchDaemons/com.canonical.multipassd.plist

# edit /var/root/Library/Application Support/multipassd/multipassd-vm-instances.json
# you'll need sudo for that

# start multipassd again
sudo launchctl load /Library/LaunchDaemons/com.canonical.multipassd.plist

How would one do this on Windows?

EDIT:

The file is here, but it doesn't appear to have an effect:

C:\ProgramData\Multipass\data\multipassd-vm-instances.json

• I stopped Multipass Service in the Windows Services app.
• I edited the file and changed mem_size to 3221225472 (3 GB).
• I started the service again.
• I launched my image and ran multipath info, but I still get Memory usage: 322.6M out of 912.5M.

Have I missed a step somewhere? Where can I find the logs?

[sebmartin]

I got it using another way. If you're using the VirtualBox driver, open the VirtualBox with the root account. There will be a new tab for Multipass with the instance already created. Just turn off the multipass instance, change the setting on the VirtualBox window and then turn on your multipass instance again.

FYI you can launch VirtualBox with the root account on macOS by running the following command in Terminal:

sudo /Applications/VirtualBox.app/Contents/MacOS/VirtualBox

[ricab]

@almontes1987, @daviewales, @sebmartin, multipass set local.<instance>.cpus=<ncpus> should be enough to get you what you need; multipass get --keys shows instance attributes that can be changed that way (along with other settings). Let us know if you have problems with that.