I am trying to do a job shop scheduling which is solved with a reinfrocement q learning agent.
This is what i got right now:
https://git.uni-wuppertal.de/1523811/tmp
Unfortunately the agent does not learn well. I think I have to overthink the state representation.
I have in that case 15 actions (1 for each Job) and also 15 Actions to place idle time (1 for each Job). In sum its 30 actions. But what is the state where the agent is? I dont know how to represent it.
Related
Since my project has so many moving parts.. probably best to explain the symptom
I have 1 scheduler running on 1 queue. I add scheduled jobs ( to be executed within seconds of the scheduling).
I keep repeating scheduling of jobs with NO rq worker doing anything (in fact, the process is completely off). In another words, the queue should just be piling up.
But ALL of a sudden.. the queue gets chopped off (randomly) and first 70-80% of jobs just disappear.
Does this have anything to do with:
the "max length" of queue? (but i dont recall seeing any limits)
does the scheduler automatically "discard" jobs where the start time
is BEFORE the current time?
ran my own experiment. RQ scheduler does indeed remove jobs whose start date < now.
I have an issue chaining AWS Batch jobs.
There are 3 Compute environments (CE_A, CE_B, CE_C) and they have associated one Job queue each (JQ_A, JQ_B, JQ_C).
There are 6 Job definitions (JD_1, JD_2, ..., JD_6).
Let <jqce>-<jd>-<name> be a Job launched on job queue (or compute environment) <jqce> and with job definition <jd>. Example: A-1-a, C-6-z.
I want to execute sequentially about 20 jobs (launched with different environment variables): A-1-a, A-1-b, B-2-c, A-3-d, A-3-e, A-3-f, ...
For each job I specify the dependency on previous job with:
params.dependsOn = [{ "jobId": "xxxxx-xxxx-xxxx-xxxxxx"}] in Batch.submitJob(params).
The first two jobs A-1-a and A-1-b execute successfully after waiting few minutes for ressource allocation.
The third job, B-2-c also executes successfully, after a some minutes of waiting for the Compute environment CE_B to be up.
Meanwhile, the compute environment CE_A is turned off since no job has presented.
HERE IS THE PROBLEM:
I expect at this point that CE_B goes down and CE_A goes up. CE_A is not going up.
The A-3-d is never executed, 16 hours later it is still in PENDING status.
The dependsOn is ok, its dependency ended long time ago.
Without dependsOn the Batch runs ok, with the same environment variables and config.
QUESTIONS
Did you face similar problems with AWS Batch and dependsOn?
Is it possible to chain batches from different Job Queues?
Is it possible to chain batches from different Compute Environments?
Does the params.dependsOn = [{ "jobId": "xxx-xxx-xxx-xxx" }] seem ok to you? It seems I do not have to set the type attribute see array jobs;
Does the params.dependsOn = [{ "jobId": "xxx-xxx-xxx-xxx" }] seem ok to you? It seems I do not have to set the type attribute see array jobs;
Yes, type is only required when it's defined as an Array job. And the JobID you're providing is what was returned when you submitted the specific job?
Is it possible to chain batches from different Job Queues?
Is it possible to chain batches from different Compute Environments?
You should be able to do it but I've never done that.
Meanwhile, the compute environment CE_A is turned off since no job has presented.
So CE_A was running already and ran A-1-a, A-1-b already?
As I recall AWS checks every 10 minutes for certain statuses and people have run into cases where the system seems stuck.
You could set CE_A to always have a minimum of 1 CPU so it doesn't disappear or become difficult to get a version of.
Can you simply for testing purposes? Shorter actions, reducing Queues, etc
Consider checking the AWS forum on Batch. Not much activity there but worth an additional set of eyes.
I have my freeRTOS currently working on my Microzed board. I am using the Xilinx SDK as the software platform and until now I have been able to create tasks and assign priority.
I was just curious to know if it would be possible to assign a fixed time for each of my tasks such that for example after 100 miliseconds my scheduler would switch to the next task . So is it possible to set a fixed execution time for each of my tasks ?? As far as I checked I could not find a method to work this out, if there is any means to implement this using the utilities of freeRTOS, kindly let me know guys.
By default FreeRTOS will time slice tasks of equal priority, see http://www.freertos.org/a00110.html#configUSE_TIME_SLICING, but there is nothing to guarantee that each task gets an equal share of the CPU. For example, interrupts use an unknown amount of processing time during each time slice, and higher priority tasks can use part or all of a time slice.
Question for you though - why would you want the behaviour you requested? Maybe if you said what you were trying to achieve, rather than than ask if a feature existed, people would be able to make helpful suggestions.
Im working on a Real world Event Management applicationwhere Events are scheduled based on availability of venue place, none of the events should clash.
What is the best possibility to insert the next value.
Example:
Suppose Event1 is happening between 9 and 9:30 at Room A
Event2 is happening between 9 and 10:00 at Room B
Event3 can start at 9:30 at Room A since its free after 9:30 n so on......
Which algorithm can I use?
Note: This is not CPU scheduling algorithm
Thanks :)
I would go with greedy approach.
Sort all the events according to start time
For each venue:
T=earliest time
Find the earliest event at or after T and schedule it for this venue. Remove that event from the list.
Update T to end time of this event
Once this is completed you can use some kind of local search (simulated annealing based) to further optimize the assignment.
I have Jenkins set up to build XBMC images for different platforms. My system takes around 6 hours to build each image, so I prefer to run them in parallel, usually 2 or 3 at a time. The problem with this is, that if they have to download updates to modules (like linux kernel or sometihng), the 2 or 3 building in parallel will download at the same time, corrupting the download (they point to the same folder)
Is it possible in jenkins/hudson to specify an offset? (I know you can schedule builds, as well as use a trigger that builds after completion of one project) something like:
Build 1: immediately
Build 2: start 20 minutes after build 1
Build 3: start 20 minutes after build 2
I tried looking for a plugin as well as google but no luck. I also know that I could schedule via the cron-like schedule capabilities in jenkins, but I have my build trigger set up to poll the GIT repo to look for changes for a build, I'm not just blind scheduling.
One way to do it is to choose the "Quiet Period" option under "Advanced".
Set it to 1200 seconds for Job 2, and 2400 seconds for Job 3.
That means Job 1 will be queued immediately when a change is noticed in git, Job 2 will go into the queue with a 20 minute delay, and Job 3 with a 40 minute delay.
Another way to do this would be to make the job some sort of a build flow (whether with the build flow plugin or by saying that the last task of job A is to run job B). If you can turn the download into its own job, then you can define the "download" job as single-threaded, and the rest as multithreaded.
Doing this serializes only what needs to be serialized. Doing an "every twenty minutes" thing will waste time when it takes fifteen minutes to download, and will fail (possibly in a hard-to-debug way) when there's a slowdown and it takes twenty-five minutes to download.