|
|
Nan Zhu 2010-10-28, 06:52
Hi, all
I'm working with Mumak recently, but I found that the code hasn't been updated for a long while, I would like to know that if the Mumak is still active? Thank you
Nan
-
Re: Mumak is still active?
Hong Tang 2010-10-29, 01:28
Nan,
We (at Yahoo) are still doing some work in and on top of Mumak in Yahoo, but not in significant ways because most of our production is using an internal version of Hadoop 20, which diverges from Hadoop trunk (where Mumak resides). So the project is certainly "live" but not very "active".
I guess you could look at it this way: would Mumak be a good base to help you accelerate what you plan to work on? If the answer is yes, take it and enhance it. Just like any open source projects, there is no particular owner of it. The more people embrace it and contribute to it, the more likely the project will be successful.
Another note I should say is that before you invest in significant amount of time in implementing something, you should create a jira ticket first, and discuss solutions with the community (sometimes even post partial patches) along the way. Sometimes your solution may not be aligned with the general design principle of the whole project, or some assumptions you make may no longer hold in future versions. Keep the community involved helps avoid such costly mistakes.
Hope this helps,
-Hong
On Oct 27, 2010, at 11:52 PM, Nan Zhu wrote:
> Hi, all > > I'm working with Mumak recently, but I found that the code hasn't been > updated for a long while, I would like to know that if the Mumak is > still > active? > > > Thank you > > Nan
-
Re: Mumak is still active?
Nan Zhu 2010-10-29, 02:04
Hi,
Hong, Thank you for the reply,
Actually I'm doing my research on Mumak before I move to a real cluster, since we are just some guys in university, so we cannot afford a large cluster containing more than 1000 nodes, I modified and enhanced the code of Mumak,
And thank you for the reminding me the efficient ways in working at a open source project, I'll create and release my solutions soon,
Here is a question about the mumak, I think it's really strange:
1. in the class of SImulatroEventQueue, the funtion of add() and addall() both limit something like this :
if (e.getTimeStamp() < lastTimeStamp) { throw new IllegalArgumentException("Event happens in the past: " + e.getClass() + "(" + e.getTimeStamp() + "<" + lastTimeStamp + " Last Event: " + lastEvent.getClass());
But I find something very strange,
I added a line of code in the createTaskCompletionAttemption in the SimulatorTaskTracker class:
LOG.info("TrackerName: " + this.taskTrackerName + "taskid:" + status.getTaskID() + "start_time:" + status.getStartTime() + "finish_time: " + status.getFinishTime());
so the program will output the information of each taskCompletionEvent,
I got something like :
--------------------------------- 10/10/29 09:48:26 INFO mapred.SimulatorTaskTracker: TrackerName: tracker_node055:localhost/127.0.0.1:10004taskid:attempt_201010132054_0019_r_000000_0start_time:1288317615942finish_time: 1288317789309 10/10/29 09:48:26 INFO mapred.SimulatorTaskTracker: TrackerName: tracker_node055:localhost/127.0.0.1:10004taskid:attempt_201010132054_0019_r_000001_0start_time:1288317618947finish_time: 1288317720289 --------------------------------- the finish-time of attempt_201010132054_0019_r_000000_0 is 1288317789309, and attempt_201010132054_0019_r_000001_0 is 1288317720289
the order shows that the first event about attempt_201010132054_0019_r_000000_0 is in front of the later one in the event queue, but the later one finishes **earlier** that the prior one, and the timestamp of completionevent is the finishtime of each event, why it hasn't throw a IllegalArgumentException I showed above?
anyone can give some explanation?
Thank you
Nan On Fri, Oct 29, 2010 at 9:28 AM, Hong Tang <[EMAIL PROTECTED]> wrote:
> Nan, > > We (at Yahoo) are still doing some work in and on top of Mumak in Yahoo, > but not in significant ways because most of our production is using an > internal version of Hadoop 20, which diverges from Hadoop trunk (where Mumak > resides). So the project is certainly "live" but not very "active". > > I guess you could look at it this way: would Mumak be a good base to help > you accelerate what you plan to work on? If the answer is yes, take it and > enhance it. Just like any open source projects, there is no particular owner > of it. The more people embrace it and contribute to it, the more likely the > project will be successful. > > Another note I should say is that before you invest in significant amount > of time in implementing something, you should create a jira ticket first, > and discuss solutions with the community (sometimes even post partial > patches) along the way. Sometimes your solution may not be aligned with the > general design principle of the whole project, or some assumptions you make > may no longer hold in future versions. Keep the community involved helps > avoid such costly mistakes. > > Hope this helps, > > -Hong > > > On Oct 27, 2010, at 11:52 PM, Nan Zhu wrote: > > Hi, all >> >> I'm working with Mumak recently, but I found that the code hasn't been >> updated for a long while, I would like to know that if the Mumak is still >> active? >> >> >> Thank you >> >> Nan >> > >
|
|