Adds option to set memory and cpu constraint

[Hephi2]

Solves #547
- Adds fields in admin UI
- Sets HostConfig.Memory and HostConfig.NanoCPUs fields for Docker Socket Proxy
- Passes memory and nanoCPU values to HaRP Proxy (requires changes in HaRP repo)

Requires changes in HaRP to work. See nextcloud/HaRP#46

To validate the effect of the resource constraints run e.g. the following commands:

• docker exec -it <container_id> bash
• Testing the cpu constraint:
  • yes > /dev/null
  • You will observe with docker stats <container_id> that the container stays around the configured CPU cap
• Testing the memory constraint:
• E.g. python3 -c "a = ' ' * (1024 * 1024 * 1024); input()"
• You will observe with docker stats <container_id> that the limit is equal to to set memory value and if the needed memory is too high, the process will be killed

[Hephi2]

Hey @kyteinsky @oleksandr-nc,
I'm not sure why the two tests are failing. I ran npm run build and there are no differences that need to be pushed.
Do you have any idea what might be causing this or what I could do?

[oleksandr-nc]

Hey @kyteinsky @oleksandr-nc, I'm not sure why the two tests are failing. I ran npm run build and there are no differences that need to be pushed. Do you have any idea what might be causing this or what I could do?

Hello! You can use /compile command to trigger the rebuild and push the assets by the bot to make CI happy

[oleksandr-nc]

P.S.: But we usually do this after the PR is approved, before merging - the only exceptions are those PRs where the author is confident there won't be any changes to not remove commits of bots from git history.

The developers are building JS locally for review.

I will try to check this PR - is it ready for review or not?

[Hephi2]

I see, that makes sense. Thank you.
Yes, it's ready for review.

[oleksandr-nc]

@Hephi2 When I edit the daemon deploy page and fill the Memory limit and then start changing the CPU limit, the Memory limit gets reset.
Can this be fixed?

[oleksandr-nc]

cpu-limit.mp4

[oleksandr-nc]

Tests were performed on HaRP, results:

Without memory limit:

root@test-deploy:/ex_app/lib# python3 test_mem.py 
Requesting 16 GiB (17,179,869,184 bytes) allocation...
Allocation succeeded, touching pages to ensure commit...
Success. Touched 16 GiB in 3.24s (pid 2406)

Result: 🆗

---

With 8GB memory limit:

root@test-deploy:/ex_app/lib# python3 test_mem.py 
Requesting 16 GiB (17,179,869,184 bytes) allocation...
Killed
root@test-deploy:/ex_app/lib# python3 test_mem.py --gib=8
Requesting 8 GiB (8,589,934,592 bytes) allocation...
Allocation succeeded, touching pages to ensure commit...
Success. Touched 8 GiB in 16.22s (pid 2009)

Result: 🆗

---

Without CPU limit:

Result: 🆗

---

With CPU limit=4:

Result: 🆗

---

@Hephi2 Second question, if you don't mind, does docker-socket-proxy require modification to support this?

[Hephi2]

@oleksandr-nc
That's strange, it works fine for me, I can't reproduce this problem.
Therefore, I can't really debug it. Perhaps you could tell me something about your configuration and the browser you are using, etc., then I might be able to reproduce it.
I can't really explain why this would happen either. Only if somehow the deployConfig object is replaced or something like that, but I don't know why this should be happening.

Regarding your second question: docker-socket-proxy does not need to be updated because, with this configuration, the Docker container is created in app_api, more specifically in lib/DeployActions/DockerActions.php.

[oleksandr-nc]

Are you using Vue2 and not building this with Vue3?
Could you please create a new Deploy Daemon and try to reproduce it by repeating the steps in the video (when initial params are not defined for memory/cpu)?

[kyteinsky]

When I edit the daemon deploy page and fill the Memory limit and then start changing the CPU limit, the Memory limit gets reset.

I can't reproduce this. Which browser do you use and does any other browser have this issue too?

[kyteinsky]

@Hephi2 Also,
MB -> MiB (mebibyte)
GB -> GiB (gebibyte)

see https://www.corsair.com/us/en/explorer/diy-builder/storage/mebibyte-vs-megabyte-whats-the-difference/
in windows mebibytes (MiB) are incorrectly labelled as MB.

[oleksandr-nc]

I can't reproduce this. Which browser do you use and does any other browser have this issue too?

Firefox. But reactivity in Vue2 should work the same among all browsers?

[oleksandr-nc]

Description what happens on video:

On mount, when editing a daemon whose resourceLimits are empty/Unlimited, the code does not create the memoryMB/cpus keys (it only converts them if existing values are truthy). So the first time you type, those keys are added after creation (non‑reactively), and the next UI update (typing in CPU) causes the Memory input to re-render using the (still empty) computed getter, resetting the field.

[DaphneMuller]

@Hephi2 would you be up for giving it one more go to see if you can reproduce it with the additional hints?

[Hephi2]

Yes, sure.
I didn't have time yesterday, but I planned to take a closer look at it today.

[kyteinsky]

When I edit the daemon deploy page and fill the Memory limit and then start changing the CPU limit, the Memory limit gets reset.
I can't reproduce this. Which browser do you use and does any other browser have this issue too?

So I was able to reproduce this issue. The steps were:

1. create a deploy daemon with no resource limit set
2. edit the daemon and add the cpu limit only
3. save it
4. edit the daemon and try to set the memory limit
5. and then try to adjust the cpu limit

the fields should be empty now.

[Hephi2]

On mount, when editing a daemon whose resourceLimits are empty/Unlimited, the code does not create the memoryMB/cpus keys (it only converts them if existing values are truthy). So the first time you type, those keys are added after creation (non‑reactively), and the next UI update (typing in CPU) causes the Memory input to re-render using the (still empty) computed getter, resetting the field.

Actually, this was covered if both were not set. However, in the edge case described by @kyteinsky, the corresponding other key was not initialised.
This should now be fixed. Please let me know if this is not the case.

[oleksandr-nc]

With latest changes editing Deploy Config works fine for me. @kyteinsky can you check on your side as well?

[kyteinsky]

confirmed, that is fixed.

[kyteinsky]

🚀

[oleksandr-nc]

/compile

[oleksandr-nc]

@Hephi2 please squash three first commits and after that we can merge this

[oleksandr-nc]

/compile