OpenKore wokring offline

[zelda1]

Colleagues,

I am curious to know if there is a virtual server on which I can host my apps while working. Because my laptop can't take all the load and be running all the time you know!!
Any thoughts??


Thanks

[Sergio]

Solving your problem is not difficult if you have some knowledge and a little experience.
Essentially, you need to rent a VPS (virtual private server) or Cloud-Instance.
Physically, the game server runs at the Hetzner DC in the Finland region. Therefore, if you want to have the shortest ping, then you should rent an Instance right away.
There are, let’s say, cheaper options, for example: digital ocean, you can look for anyone else... in principle, a ping of 150 milliseconds will not be a significant problem for the bot. Choosing a DC is not a big problem! I don’t recommend hiring from big DCs like Google or Amazon - they require a lot of specific knowledge and it will cost much more!
The problem is how you are going to use the Instance. In order to run the bot, the server must have an OS, for example: Windows or Linux.
Linux requires a lot of knowledge and experience.
To begin with, you will need to install Windows OS on the Instance, but then you will have additional costs for the OS license! https://youtu.be/yzBsWd23Nw4 - here is an example of a video on how you can do a trick on a Hetzner instance - a man shows that you can buy a Linux instance, then reconfigure and install Windows - in this case there will be no license, but Windows will be require activation. This does not violate the rules of operation of the instance and Windows!
This is a good option. because without Activation, Windows will only connect one user and will not provide the opportunity to customize the desktop (you don’t need to do this on the Instance either)
The next step: you need to configure RDP (Remote desktop server) on the instance and then you can connect to your Virtual Server through the mstsc (MicroSoft terminal server client) program - and launch bots, just like on your laptop.

As an example, I can give you that in Hetzner on a CX21 instance - with Linux in console mode I can use approximately 10-12 bots. I think that on Windows you could run 8 bots in console mode, if you run bots with a map then 4-5. I'm jealous of how to set everything up... in Windows, at a minimum, you still need to use an antivirus, I like the AVG-free version, you can use your favorite one. Well, sometimes you need to use a Web browser there... i.e. the instance must have at least 4GB of RAM for normal operation.

I don't know how difficult this will all be for you. These are quite specific skills of the DevOps/SysAdmin professions. For us, this is both a job and a hobby, if we really want to, we can carry out such work and teach how to do it ourselves, but this will cost money

[zelda1]

well, this is so informative and concise I would say. thanks a lot.
I will give it a try and see how it goes!!

[zelda1]

Hi Sergio, so I managed to get them to work on CX21 with Win installed. however, I started receiving a bug issue (busparty line bug issue) from the bot also I see my bots are sleeping when left for a long time and they reconnect as soon as I open the server again. any idea on how to fix these issues?

[Sergio]

Congratulations!
Now do it right:
1. If you have an error, create a topic for it in the forum. And then we will discuss and solve it.
2. attach a screenshot or text of the error (text is better) and examples of your configs.
3. if you solved the problem, then edit the topic - indicate the solution or the answers that solved the problem.
And then your work will not be in vain - you will begin to truly enjoy our server

[Sergio]

Solved & Closed

[Sergio]

Hello, folks!
I've encountered an issue with Hetzner cloud instances. I've been struggling with it for almost 2 months, but I've grasped the essence of it. As part of my work, I have numerous instances in both Hetzner and Amazon. Amazon strictly specifies that instances should not be overloaded – otherwise, they throttle resources. They have various metrics for this, developed by Amazon themselves. But Hetzner also restricts CPU load!
Since I use instances for work tasks, and OpenKore is only additional load, I can't provide more precise numbers for understanding. I can only explain the principle.
Hetzner's Shared vCPU in Finland restricts at the level of 66-75%. Everything depends on the type of load. For example, I use CX21 and CX31. These instances have 2 vCPUs each. I could run 8 uOK207 with minimal logic, and the instance would operate with the main task – and this all had a load of less than 66%, so I assumed the instance could handle 12 instances of OpenKore 2.0.7.
Currently, even 4 bots of uOK210 with stricter timings, bus-server, in a base party, and extended logic through macros occasionally overload the instance's processor.
Based on practice, OpenKore 2.1.0 consumes at least 10-20% more CPU compared to OpenKore 2.0.7.
Highly recommended:

    If you use a solo bot, use uOK207.
    Don't overload OpenKore with very small timeout values. I know these values are greatly exaggerated, and the bot works very sluggishly as a result. But in most cases, using a timeout of less than 1 second will be very disadvantageous in terms of system resource load and OpenKore AI operation.
    Use a minimum number of plugins.
    Carefully optimize macros – use timeouts and delays, and increase their value where possible.
    Optimize the main config.txt – disable unnecessary things!
    Now, I'll describe how to understand if the instance is subject to CPU restrictions.
    Working with the instance is difficult – everything runs slowly and glitches heavily.
    The OS in the instance shows a total load of ~100%.
    The CPU graph in the Hetzner console shows less than 100% load!
    The CPU graph in the Hetzner console is a very tricky tool, you must understand it! We're used to understanding that 100% is the maximum value. But Hetzner shows it differently, it all depends on the number of CPUs in the instance. If you buy an instance with 4 CPUs, then the maximum load in the graph will be 400%. And if 8 CPUs, then 800% - accordingly!
    On my instances, the maximum load can be 200% (2 CPUs), but when Hetzner imposes restrictions, the graph shows ~50%. That is, a decrease by 4 times!
    I don't know the situation with Dedicated vCPU.
    In other regions, the situation may be better, I don't know the details yet. 4-5 years ago, in some region of Germany, I observed a situation: if the instance was loaded >80% cumulatively for all CPUs for 4 hours, then Hetzner simply rebooted the instance.

---

Привет, народ!
Выловил проблему в облачных инстансах Хетзнера. Боролся почти 2 месяца с проблемой, но суть я понял. По своей работе я имею много инстансов в Хетзнере и Амазон. Амазон строго оговаривает, что нельзя перегружать инстансы – иначе они ограничивают ресурсы. Для этого есть разные показатели, разработанные самим Амазон. Но Хетзнер тоже ограничивает загрузку цпу!
Т.к. я использую инстансы под рабочие задачи, а OpenKore - это только дополнительная нагрузка, то не могу привести более точных цифр для понимания. Могу объяснить только принцип.
Shared vCPU Хетзнер в Финляндии ограничивает на уровне 66-75%. Всё зависит от типа нагрузки. На пример, я пользуюсь CX21 и CX31. Эти инстансы имеют по 2  vCPU. Я мог запустить 8 uOK207 с минимальной логикой и на инстансе работала основная задача – и это всё имело нагрузку менее 66%, поэтому я и предполагал, что инстанс может работать с 12 экземплярами OpenKore 2.0.7.
На текущий момент даже 4 бота uOK210 с более жесткими таймингами, шиной-сервером, в бас-пати, и расширенной логикой через макросы периодически перегружают процессор инстанса.
Исходя из практики, то OpenKore 2.1.0 потребляет минимально на 10-20% больше цпу по сравнению от OpenKore 2.0.7.
Очень рекомендую:
1. если вы используете соло-бота, то используйте uOK207.
2. Не перегружайте OpenKore очень маленькими значениями таймаутов. Я знаю, что эти значения сильно завышены, и бот работает очень тупо поэтому. Но в большинстве случаев использовать тайминг меньше 1 секунды будет очень невыгодно из расчета нагрузки на ресурсы системы и работы ИИ OpenKore.
3. Используйте минимум плагинов.
4. Внимательно оптимизируйте макросы – используйте таймауты и задержки, и увеличивайте их значение при возможности.
5. Оптимизируйте основной config.txt – отключайте не нужное!
Сейчас опишу, как понять, что на инстанс действуют ограничения по ЦПУ.
1. Работать с инстансом тяжело – всё работает медленно и сильно глючит
2. ОС в инстансе показывает общую загрузку ~100%
3. График ЦПУ в консоле Хетзнера показывает меньше 100% нагрузку!
График ЦПУ в консоли Хетзнера – это очень хитрая утилита, её нужно обязательно понять! Мы привыкли понимать, что 100% это максимальная величина. Но Хетзнер показывает по-другому, всё зависит от количества ЦПУ в инстансе. Если купить инстанс с 4 ЦПУ, то максимальная нагрузка в графике будет 400%. А если 8 цпу, то 800% - соответственно!
У меня на моих инстансах максимальная нагрузка может быть 200% (2 цпу), но когда Хетзнер накладывает ограничения, то график показывает ~50%. Т.е. понижение в 4 раза!
Ситуацию с Dedicated vCPU я не знаю.
В других регионах, возможно, ситуация лучше, подробнее пока не знаю. 4-5 лет назад, в каком-то регионе Германии наблюдал ситуацию: если нагружать инстанс >80% суммарно по всем цпу 4 часа, то хетзнер просто перезагружали инстанс.

[Sergio]

Выявил такую проблему с Хетзнером:
Не так давно - меньше недели назад, эти ребята стали совсем не друзьями! Раньше инстанс можно было грузить на 3/4 заявленной мощности по цпу, то сейчас всеголишь до 1/2. И это при том, что несколько месяцев подряд они повышали тарифы.
В ообщем понимании - Хетзнер Клоуд, стал не самым лучшим решением в отношении цена/качество.

Заново открываю данную дискусию - давайте скать вместе варианты микроинстанов для РО-ботоводства!
----
I discovered the following problem with Hetzner:
Not so long ago - less than a week ago, these guys became not friends at all! Previously, an instance could be loaded at 3/4 of the declared CPU capacity, but now it’s only up to 1/2. And this despite the fact that for several months in a row they raised tariffs.
In a general sense, Hetzner Cloud was not the best solution in terms of price/quality.

I’m re-opening this discussion - let’s find together options for micro-installations for RO-bot farming!

[zelda1]

Thanks, Sergio for your message, I agree with you. I was kinda forced to upgrade my server to the newly released CX32 as the the old ones were lagging and eventually got hacked. This is the way they force their clients to upgrade without telling them...
I will look around for alternatives and will let you know