Add connection pooling

[dsully]

Any interest in merging pooled connection support if I were to implement it?

Thanks

[earl]

In principle, yes. Even more so if it can be kept as simple as possible. Could you shortly describe how you envision it behaving, in rough terms?

[dsully]

I was looking at doing a rough port of how the Ruby client works - some operations read from all hosts in the pool & collate the results, other operations work against a specific host.

I also prefer simple - and am thinking of a separate class BeanstalkPool or similar, but there might need to be some changes to the core code.

Thanks

[earl]

I assume you are talking of beaneater? I'll have a look again at how they do pooling.

[dsully]

Yes, beaneater.

[Doerge]

Hi there.

As part of my job I forked and wrote a simple Pool-class. I haven't looked at beaneater, but it sounds like I'm doing exactly what you talk about. I haven't made a pull-request yet because I thought we would come across various issues while coding up against the pool, and I have managed to catch a bug or two this way. I will make a pull-request when I feel it is deserves to be a part of this great client!

[ysmolski]

I think having pool is essential for such a client. I have noticed that Doerge has made additions with Pool class [1]. Earl, what do you think about merging into your original library?

1. https://github.com/Doerge/beanstalkc

[ysmolski]

Doerge, do you mind if I work on your fork with Pooled connections to cut some rough edges? :-)

[earl]

If someone prepares a pull-request, I'll be happy to review it. Doerge mentioned that he came across various issues he wanted to shake out first; I don't know what the status of that is.

[Doerge]

Hiya.

We are using my branch in production and it has been a while since I came across any issues. I don't know if that justifies a pull-request? (I'm new to github'ish..)

@ysmolsky Fork away! I would love to see some polish! For the sake of my coding-bettering what rough edges have you noticed?

edit:
I logged 3 issues with the pool btw:
https://github.com/Doerge/beanstalkc/issues?state=open

[ysmolski]

Well, the polish I meant is about code style mostly. But more importantly is that reconnection does not work as expected for your fork, b/c of issues you mentioned by that link.

For example the most obvious is the we need to reuse and rewatch all those tubes we were using and watching prior the disconnect. I am going to implement them in fork with BlockingConnection class...

Also I don't really like the idea of having multiple beastalkd server and pulling reserve() from random instances. Thinking about having something more reliable..

[Doerge]

Cool! I'm looking forward to learn from your polishing!

@reconnecting
Cool! I'll keep an eye on your fork :)

@reserving from random instances:
Yeah I did consider peeking at the different bstalkds but for our scenario I went with the simple solution at first. Setting a time out on the reserve and then trying the next in line should be a better solution.