Skip to content
Commits on May 31, 2011
  1. changed the socket name in the worker, also just ran a test of runnin…

    committed May 31, 2011
    …g multiple times and it worked, there is no requirement other than telling it what to connect to.
  2. dropped the identity on the sub socket, not necessary. Working on dro…

    committed May 31, 2011
    …pping the need for one on the broker_socket as well. Speaking of which I still need to rename that socket
  3. ok, figured out the problem had to do with the pubs I was doing, kind…

    committed May 30, 2011
    …a dumb mistake. Make some even sillier ones fixing it like forgetting a list index starts with 0, think I'm really tired and should sleep. Pretty excited though because I think I'm closing on being able to work on the client side soon
  4. got timestamps working, however sending messages using the worker_id …

    committed May 30, 2011
    …as the subscription key isn't working. Also had a weird issue where worked_id wasn't acceptable as a key unless I checked to see if it was there using has_key first or not. Somewhere something is broken, not sure where yet though
Commits on May 29, 2011
  1. got pub/sub working now on the side. Think I figured out the issue, h…

    committed May 29, 2011
    …opefully can get some time this weekend to move to pub/sub for getting messages to workers soon. Also slowly working on getting socket names less confusing by naming the socket by the app that defines it rather than the app it's connecting to. Last, since UUID uses the - character, will be moving to _ in socket names as a delimter as well as just getting the UUID in the LRU/worker queue definitions rather the full socket name. Well that's a wordy commit
  2. tweaked the way heartbeats are sent a bit and now persist socket, whi…

    committed May 29, 2011
    …ch should make client requests a little fast. Laptop battery dieing so calling it a night on this
Commits on May 26, 2011
  1. Edited the README to include information on alternatives to Scale0 in…

    committed May 26, 2011
    … standard deployments and a new note on how Scale0 will avoid being a single point of failure. (oh and I love I can do this through the github web interface)
  2. Updated thre readme, also moved playground to Scale0, going to start …

    committed May 25, 2011
    …making this a proper python project after I can get a full frontend to multiple backend example up.
Commits on May 23, 2011
  1. Merge pull request #1 from minrk/zmqstream

    committed May 23, 2011
    use zmqstream instead of pure sockets
  2. @minrk
  3. moving to a different socket didn't help, commiting to github, going …

    committed May 22, 2011
    …to see if I can get help from the list figuring out what I'm doing wrong.
Commits on May 22, 2011
  1. getting dropped pings, not sure why. First going to test on a rackspa…

    committed May 22, 2011
    …ce server which has more cores
  2. ping/pong scale0 to worker implemented. Also move the LRU to be a dic…

    committed May 22, 2011
    …t with the key being the worker_id for faster access to modifying the timestamp. As I commit, I realize this ruins the whole LRU concept. Might have to break this into a dict for information and a list for the actual queue. In the end this could give me a place to put some sort of statistics though, so may not be a bad idea
  3. basic worker to scale0 heartbeat in place, think I'm going to change …

    committed May 22, 2011
    …it though. Will come back to it after I get LRU queue management by putting scale0 to worker heartbeat in
Commits on May 15, 2011
  1. ping pong success\! Funny how you have to write code to get some conc…

    committed May 15, 2011
    …epts to stick. Have to include the id of the recipient to send messages via a XREP (or ROUTER) socket. You know, like the guide says
Commits on May 14, 2011
  1. maybe a lesson learned if I do a blog post. zmq.POLLIN only carries o…

    committed May 14, 2011
    …n to the next loop when a connection has been made. Pretty sweet actually. Anyway, I've dumbed it way down, right now worker PINGs broker, router multiprocess concept is completely gone, for now. For some reason it appeared to be breaking polling. I think I just need to build a little slower. So concentrating on ping/pong now. Client down to Worker will be next, and then I'll try to plug the router objects in next. Honestly, I think I'm battling Python right now, the GIL sucks. Thinking with something like erlang I could thread without traditional threading resource consumption and still take advtange of multiple processors
  2. having some problems with the poller, looks like it's hanging on the …

    committed May 13, 2011
    …poller.poll(), but not sure why. Getting late and I'm tired. Will look at it more later. Github push to not lose work
Commits on May 13, 2011
  1. going ahead and plugging in router as I go, would have to tear the ot…

    committed May 13, 2011
    …her method skipping it apart to do it, but have to quit for now so getting a commit and push to github in, just in case I work on this with another machine
Commits on May 12, 2011
  1. actually diving in and writing code, I realized I was doing it wrong.…

    committed May 12, 2011
    … Workers will provide the Broker a socket the Broker can connect to it on. The Dispatcher will pass that to the Router, which will make a REQ to the worker itself. The Worker will need to support a socket for communication with the Broker, and another for actually parsing requests
  2. setting up the playground directory, taking Zed Shaw's advice and try…

    committed May 12, 2011
    …ing to just knock something out in Python to get a better feel of the actual problem
Commits on May 11, 2011
Commits on Apr 25, 2011
  1. Edited README via GitHub

    committed Apr 25, 2011
Commits on Apr 22, 2011
  1. starting on Dispatcher. honestly just can't focus on this now. Trying…

    committed Apr 22, 2011
    … to program while refreshing web pages and digging through snort/base every few minutes as part of my job just plain doesn't work. Was hoping to have to something by the weekend, but looks like that's not going to happen
  2. Worker class started

    committed Apr 22, 2011
Commits on Apr 21, 2011
  1. fixing a filename

    committed Apr 21, 2011
  2. oops. Never read the pyzmq docs, duh. Just realized that I was jumpin…

    committed Apr 21, 2011
    …g into the examples. Stopping in order to read the docs before I really start trying to put anything together.
  3. ok still moving around how I'm thinknig about this. The Dispatcher/Ro…

    committed Apr 21, 2011
    …uter communication really only needs to be one socket. A request is used to put it in the LRU Queue, and a request can be used to send a reply.
Something went wrong with that request. Please try again.