Lets say I have 5 GB Input File and I have Cluster Setup of 3 Data Nodes with each 25 cores (Total - 75 cores) and 72GB memory (Total - 216GB Memory).
How to calculate number of executors, number of cores and executor memory for this particular file size and memory configuration.
How many blocks will create in HDFS for this file?
Method for executor resource calculation:
Allocate 1 core and 1 GB per node for yarn. So, we are left with 72 core and 213 GB memory.
The remaining resource have ~1:2.95 core to (GB) memory ratio.
The optimal CPU count per executor is 5. So, to prevent underutilisation of CPU or memory resource, the executor’s optimal resource per executor will be 14.7GB(5*2.95) memory and 5 CPU.
You should keep block size as 128MB and use same as spark parameter:
spark.sql.files.maxPartitionBytes=134217728.
It will result in 40 blocks or partitions.
You can refer this article for more details. https://devendraparhate.medium.com/apache-spark-job-aws-emr-cluster-s3-yarn-and-hdfs-tuning-77514afb9ce8
My Spark job failed with the YARN error Container killed by YARN for exceeding memory limits 10.0 GB of 10 GB physical memory used.
Intuitively, I decreased the number of cores from 5 to 1 and the job ran successfully.
I did not increase the executor-memory because 10g was the max for my YARN cluster.
I just wanted to confirm if my intuition. Does reducing the number of executor-cores consume less executor-memory? If so, why?
spark.executor.cores = 5, spark.executor.memory=10G
This means an executor can run 5 tasks in parallel. That means 10 GB needs to be shared by 5 tasks.So effectively on an average - each task will have 2 GB available. If all the tasks consumes more than 2 GB, than overall JVM will end up consuming more than 10 GB and so YARN will kill the container.
spark.executor.cores = 1, spark.executor.memory=10G
This means an executor can run only 1 task. That means 10 GB is available to 1 task completely. So if the task uses more than 2 GB but less than 10 GB, it will work fine. That was the case in your Job and so it worked.
Yes, each executor uses an extra 7% of memoryOverhead.
This calculation will be created thinking that you have two nodes, so we have three executors in one node and two executors in the other node.
Memory per executor in the first node = 10GB/3 = 3,333GB
Counting off heap overhead = 7% of 3,333GB = 0,233GB.
So, your executor-memory should be 3,333GB - 0,233GB = 3,1GB per node
You can read another explanation here:
https://spoddutur.github.io/spark-notes/distribution_of_executors_cores_and_memory_for_spark_application.html
I have a question regarding spark. I am using spark 2.2 and as per my knowledge each executor spins up taks and executes the task. Each task corresponds to a partition. Default number of partition is based on default parallelism and the file size/Default Block Size. So considering a file size of 1 GB and a cluster of 4 executors each of which can spin up 2 tasks (2 core). As per calculation the executor memory should be about 256 MB (2 tasks each task operating on 128 MB block)+ 384 MB overhead. However If I run the code with this size as executor memory the performance is slow. If I give executor memory of 1.5 GB (considering some calculations on rdd) still the performance is slow. Only when I increase the executor memory to 3GB the performance is Good.
Can someone explain
1. why do we need so much executor memory when we work on only 128 MB of data at a time.
2. How do we calculate the optimum executor memory needed for the job
Thanks for your help
I have a cluster of 5 MemSQL Nodes and 2 Aggregators running this on 300 GB of memory.
Our size of row stores is fairly small less than 20 GB to be precise. The problem that we are facing is after few days the memsql memory is at 100% and at that time the application has to be restarted.
Is there a way to force the memory clean up
Where do you start to tune the above mentioned params. Do we start with executor memory and get number of executors, or we start with cores and get the executor number. I followed the link. However got a high level idea, but still not sure how or where to start and arrive to a final conclusion.
The following answer covers the 3 main aspects mentioned in title - number of executors, executor memory and number of cores. There may be other parameters like driver memory and others which I did not address as of this answer, but would like to add in near future.
Case 1 Hardware - 6 Nodes, and Each node 16 cores, 64 GB RAM
Each executor is a JVM instance. So we can have multiple executors in a single Node
First 1 core and 1 GB is needed for OS and Hadoop Daemons, so available are 15 cores, 63 GB RAM for each node
Start with how to choose number of cores:
Number of cores = Concurrent tasks as executor can run
So we might think, more concurrent tasks for each executor will give better performance. But research shows that
any application with more than 5 concurrent tasks, would lead to bad show. So stick this to 5.
This number came from the ability of executor and not from how many cores a system has. So the number 5 stays same
even if you have double(32) cores in the CPU.
Number of executors:
Coming back to next step, with 5 as cores per executor, and 15 as total available cores in one Node(CPU) - we come to
3 executors per node.
So with 6 nodes, and 3 executors per node - we get 18 executors. Out of 18 we need 1 executor (java process) for AM in YARN we get 17 executors
This 17 is the number we give to spark using --num-executors while running from spark-submit shell command
Memory for each executor:
From above step, we have 3 executors per node. And available RAM is 63 GB
So memory for each executor is 63/3 = 21GB.
However small overhead memory is also needed to determine the full memory request to YARN for each executor.
Formula for that over head is max(384, .07 * spark.executor.memory)
Calculating that overhead - .07 * 21 (Here 21 is calculated as above 63/3)
= 1.47
Since 1.47 GB > 384 MB, the over head is 1.47.
Take the above from each 21 above => 21 - 1.47 ~ 19 GB
So executor memory - 19 GB
Final numbers - Executors - 17, Cores 5, Executor Memory - 19 GB
Case 2 Hardware : Same 6 Node, 32 Cores, 64 GB
5 is same for good concurrency
Number of executors for each node = 32/5 ~ 6
So total executors = 6 * 6 Nodes = 36. Then final number is 36 - 1 for AM = 35
Executor memory is : 6 executors for each node. 63/6 ~ 10 . Over head is .07 * 10 = 700 MB. So rounding to 1GB as over head, we get 10-1 = 9 GB
Final numbers - Executors - 35, Cores 5, Executor Memory - 9 GB
Case 3
The above scenarios start with accepting number of cores as fixed and moving to # of executors and memory.
Now for first case, if we think we dont need 19 GB, and just 10 GB is sufficient, then following are the numbers:
cores 5
# of executors for each node = 3
At this stage, this would lead to 21, and then 19 as per our first calculation. But since we thought 10 is ok (assume little overhead), then we cant switch # of executors
per node to 6 (like 63/10). Becase with 6 executors per node and 5 cores it comes down to 30 cores per node, when we only have 16 cores. So we also need to change number of
cores for each executor.
So calculating again,
The magic number 5 comes to 3 (any number less than or equal to 5). So with 3 cores, and 15 available cores - we get 5 executors per node. So (5*6 -1) = 29 executors
So memory is 63/5 ~ 12. Over head is 12*.07=.84
So executor memory is 12 - 1 GB = 11 GB
Final Numbers are 29 executors, 3 cores, executor memory is 11 GB
Dynamic Allocation:
Note : Upper bound for the number of executors if dynamic allocation is enabled. So this says that spark application can eat away all the resources if needed. So in
a cluster where you have other applications are running and they also need cores to run the tasks, please make sure you do it at cluster level. I mean you can allocate
specific number of cores for YARN based on user access. So you can create spark_user may be and then give cores (min/max) for that user. These limits are for sharing between spark and other applications which run on YARN.
spark.dynamicAllocation.enabled - When this is set to true - We need not mention executors. The reason is below:
The static params number we give at spark-submit is for the entire job duration. However if dynamic allocation comes into picture, there would be different stages like
What to start with :
Initial number of executors (spark.dynamicAllocation.initialExecutors) to start with
How many :
Then based on load (tasks pending) how many to request. This would eventually be the numbers what we give at spark-submit in static way. So once the initial executor numbers are set, we go to min (spark.dynamicAllocation.minExecutors) and max (spark.dynamicAllocation.maxExecutors) numbers.
When to ask or give:
When do we request new executors (spark.dynamicAllocation.schedulerBacklogTimeout) - There have been pending tasks for this much duration. so request. number of executors requested in each round increases exponentially from the previous round. For instance, an application will add 1 executor in the first round, and then 2, 4, 8 and so on executors in the subsequent rounds. At a specific point, the above max comes into picture
when do we give away an executor (spark.dynamicAllocation.executorIdleTimeout) -
Please correct me if I missed anything. The above is my understanding based on the blog i shared in question and some online resources. Thank you.
References:
http://site.clairvoyantsoft.com/understanding-resource-allocation-configurations-spark-application/
http://spark.apache.org/docs/latest/configuration.html#dynamic-allocation
http://spark.apache.org/docs/latest/job-scheduling.html#resource-allocation-policy
Also, it depends on your use case, an important config parameter is:
spark.memory.fraction(Fraction of (heap space - 300MB) used for execution and storage) from http://spark.apache.org/docs/latest/configuration.html#memory-management.
If you dont use cache/persist, set it to 0.1 so you have all the memory for your program.
If you use cache/persist, you can check the memory taken by:
sc.getExecutorMemoryStatus.map(a => (a._2._1 - a._2._2)/(1024.0*1024*1024)).sum
Do you read data from HDFS or from HTTP?
Again, a tuning depend on your use case.