This is not like proving something that does not exist, like claiming there is an elephant in the building and it is living in a safe and trying to disprove it. It can be true or not, but in order to disprove it you need to check all the safes. This is not the case here. If you have a stateful clusters, all you need to do is do load test and crush one of the nodes during load test. This is not a philosophy question.
I should have been more clear on that entry. It is not directly related to stateful clusters. What I meant by that was:
Before executing a step, Integration Server passes the pipeline data to that step.
Step executes with that pipeline data, and creates another pipeline after executing that. If the step involves multiple inner steps, same process will be applied there as well.
In short, every step executed generates a new pipeline. I donât know how frequent integration server saves the pipeline. But stateful clusters will use the database to save the pipeline. It may be creating check points for every step, or it may be creating random check points. It can only be known by the developers themselves, unless they disclose that information somewhere, like documentation. Without that information what I did was only a speculation, a guess.
Referencing a past POC data is OK, but claiming something strongly just because you have more experience then someone else is not scientific. According to that logic oldest people should know everything. Its certainly not the case, as people grow older they tend to reject learning new things and they want to keep doing the same thing they have been doing the most. Thats why kids are really better with technology then most of us. This doesnât indicate that as people grow older then donât learn anything or they learn less. It is not, experience is important. But when there is a disagreement on a subject we need to stick with the facts. Thatâs how science makes progress.
Exactly my point. How do you know it is still the same? Windows 95 runs on DOS. According to my past 30 years of windows experience can I claim windows still runs on DOS? People are still disabling swap space because they think it slows down the execution speed by using page file. May be it was the case when we had 64 mb of rams and for servers it wasnât necessary. It certainly is not the case anymore. Same thing applies to using nolock in every tsql select query, or running services as transformers. We should update our knowledge and believes frequently. I asked that question to ask you all when was the last time you updated your knowledge on this subject.
My point is, if stateful clustering doesnât do anything for reliability then, why implement it? What is the use case for it? Why add it to documentation? It canât work without F5 anyway, what is the point having it according to you all? What do you think Devs were thinking when they implemented it?
I didnât reply unrelated parts in order not to get off topic. I am not talking about check point pattern. Stateful clusters may or may not use that out of the box. I donât have that information about it so no need to go off topic here.
If I sound rude or something, please excuse me. English is not my primary language, and I certainly am not too polite even when speaking my primary language. That certainly is not intentional. Just wanted to clarify it just in case.
