I am running an Aerospike cluster in Google Cloud. Following the recommendation on this post, I updated to the last version (3.11.1.1) and re-created all servers. In fact, this change cause my 5 servers to operate in a much lower CPU load (it was around 75% load before, now it is on 20%, as show in the graph bellow:
Because of this low load, I decided to reduce the cluster size to 4 servers. When I did this, my application started to receive the following error:
All batch queues are full
I found this discussion about the topic, recommending to change the parameters batch-index-threads and batch-max-unused-buffers with the command
asadm -e "asinfo -v 'set-config:context=service;batch-index-threads=NEW_VALUE'"
I tried many combinations of values (batch-index-threads with 2,4,8,16) and none of them solved the problem, and also changing the batch-index-threads param. Nothing solves my problem. I keep receiving the All batch queues are full error.
Here is my aerospace.conf relevant information:
service {
user root
group root
paxos-single-replica-limit 1 # Number of nodes where the replica count is automatically reduced to 1.
paxos-recovery-policy auto-reset-master
pidfile /var/run/aerospike/asd.pid
service-threads 32
transaction-queues 32
transaction-threads-per-queue 4
batch-index-threads 40
proto-fd-max 15000
batch-max-requests 30000
replication-fire-and-forget true
}
I use 300GB SSD disks on these servers.
A quick note which may or may not pertain to you:
A common mistake we have seen in the past is that developers decide to use 'batch get' as a general purpose 'get' for single and multiple record requests. The single record get will perform better for single record requests.
It's possible that you are being constrained by the network between the clients and servers. Reducing from 5 to 4 nodes reduced the aggregate pipe. In addition, removing a node will start cluster migrations which adds additional network load.
I would look at the batch-max-buffer-per-queue config parameter.
Maximum number of 128KB response buffers allowed in each batch index
queue. If all batch index queues are full, new batch requests are
rejected.
In conjunction with raising this value from the default of 255 you will want to also raise the batch-max-unused-buffers to batch-index-threads x batch-max-buffer-per-queue + 1 (at least). If you do not do that new buffers will be created and destroyed constantly, as the amount of free (unused) buffers is smaller than the ones you're using. The moment the batch response is served the system will strive to trim the buffers down to the max unused number. You will see this reflected in the batch_index_created_buffers metric constantly rising.
Be aware that you need to have enough DRAM for this. For example if you raise the batch-max-buffer-per-queue to 320 you will consume
40 (`batch-index-threads`) x 320 (`batch-max-buffer-per-queue`) x 128K = 1600MB
For the sake of performance the batch-max-unused-buffers should be set to 13000 which will have a max memory consumption of 1625MB (1.59GB) per-node.
Related
We have a Vertex AI model that takes a relatively long time to return a prediction.
When hitting the model endpoint with one instance, things work fine. But batch jobs of size say 1000 instances end up with around 150 504 errors (upstream request timeout). (We actually need to send batches of 65K but I'm troubleshooting with 1000).
I tried increasing the number of replicas assuming that the # of instances handed to the model would be (1000/# of replicas) but that doesn't seem to be the case.
I then read that the default batch size is 64 and so tried decreasing the batch size to 4 like this from the python code that creates the batch job:
model_parameters = dict(batch_size=4)
def run_batch_prediction_job(vertex_config):
aiplatform.init(
project=vertex_config.vertex_project, location=vertex_config.location
)
model = aiplatform.Model(vertex_config.model_resource_name)
model_params = dict(batch_size=4)
batch_params = dict(
job_display_name=vertex_config.job_display_name,
gcs_source=vertex_config.gcs_source,
gcs_destination_prefix=vertex_config.gcs_destination,
machine_type=vertex_config.machine_type,
accelerator_count=vertex_config.accelerator_count,
accelerator_type=vertex_config.accelerator_type,
starting_replica_count=replica_count,
max_replica_count=replica_count,
sync=vertex_config.sync,
model_parameters=model_params
)
batch_prediction_job = model.batch_predict(**batch_params)
batch_prediction_job.wait()
return batch_prediction_job
I've also tried increasing the machine type to n1-high-cpu-16 and that helped somewhat but I'm not sure I understand how batches are sent to replicas?
Is there another way to decrease the number of instances sent to the model?
Or is there a way to increase the timeout?
Is there log output I can use to help figure this out?
Thanks
Answering your follow up question above.
Is that timeout for a single instance request or a batch request. Also, is it in seconds?
This is a timeout for the batch job creation request.
The timeout is in seconds, according to create_batch_prediction_job() timeout refers to rpc timeout. If we trace the code we will end up here and eventually to gapic where timeout is properly described.
timeout (float): The amount of time in seconds to wait for the RPC
to complete. Note that if ``retry`` is used, this timeout
applies to each individual attempt and the overall time it
takes for this method to complete may be longer. If
unspecified, the the default timeout in the client
configuration is used. If ``None``, then the RPC method will
not time out.
What I could suggest is to stick with whatever is working for your prediction model. If ever adding the timeout will improve your model might as well build on it along with your initial solution where you used a machine with a higher spec. You can also try using a machine with higher memory like the n1-highmem-* family.
I am using AWS Redis for a project and ran into an Out of Memory (OOM) issue. In investigating the issue, I discovered a couple parameters that affect the amount of usable memory, but the math doesn't seem to work out for my case. Am I missing any variables?
I'm using:
3 shards, 3 nodes per shard
cache.t2.micro instance type
default.redis4.0.cluster.on cache parameter group
The ElastiCache website says cache.t2.micro has 0.555 GiB = 0.555 * 2^30 B = 595,926,712 B memory.
default.redis4.0.cluster.on parameter group has maxmemory = 581,959,680 (just under the instance memory) and reserved-memory-percent = 25%. 581,959,680 B * 0.75 = 436,469,760 B available.
Now, looking at the BytesUsedForCache metric in CloudWatch when I ran out of memory, I see nodes around 457M, 437M, 397M, 393M bytes. It shouldn't be possible for a node to be above the 436M bytes calculated above!
What am I missing; Is there something else that determines how much memory is usable?
I remember reading it somewhere but I can not find it right now. I believe BytesUsedForCache is a sum of RAM and SWAP used by Redis to store data/buffers.
As Elasticache's docs suggest that SWAP should not go higher than 300 MB.
I would suggest checking the SWAP metric at that time.
I am trying to determine how many nodes I need for my EMR cluster. As part of best practices the recommendations are:
(Total Mappers needed for your job + Time taken to process) / (per instance capacity + desired time) as outlined here: http://www.slideshare.net/AmazonWebServices/amazon-elastic-mapreduce-deep-dive-and-best-practices-bdt404-aws-reinvent-2013, page 89.
The question is how to determine how many parallel mappers the instance will support since AWS don't publish? https://aws.amazon.com/emr/pricing/
Sorry if i missed something obvious.
Wayne
To determine the number of parallel mappers , you will need to check this documentation from EMR called Task Configuration where EMR had a predefined mapping set of configurations for every instance type which would determine the number of mappers/reducers.
http://docs.aws.amazon.com/emr/latest/ReleaseGuide/emr-hadoop-task-config.html
For example : Lets say you have 5 m1.xlarge core nodes. According to the default mapred-site.xml configuration values for that instance type from EMR docs, we have
mapreduce.map.memory.mb = 768
yarn.nodemanager.resource.memory-mb = 12288
yarn.scheduler.maximum-allocation-mb = 12288 (same as above)
You can simply divide the later with former setting to get the maximum number of mappers supported by one m1.xlarge node = (12288/768) = 16
So, for the 5 node cluster , it would a max of 16*5 = 80 mappers that can run in parallel (considering a map only job). The same is the case with max parallel Reducers(30). You can do similar math for a combination of mappers and reducers.
So, If you want to run more mappers in parallel , you can either re-size the cluster or reduce the mapreduce.map.memory.mb(and its heap mapreduce.map.java.opts) on every node and restart NM to
To understand what the above mapred-site.xml properties mean and why you do need to do those calculations , you can refer it here :
https://hadoop.apache.org/docs/r2.7.2/hadoop-yarn/hadoop-yarn-common/yarn-default.xml
Note : The above calculations and statements are true if EMR stays in its default configuration using YARN capacity scheduler with DefaultResourceCalculator. If for example , you configure your capacity scheduler to use DominantResourceCalculator, it will consider VCPU's + Memory on every nodes (not just memory's) to decide on parallel number of mappers.
Referring to the docs, you can specify the number of concurrent connection when pushing large files to Amazon Web Services s3 using the multipart uploader. While it does say the concurrency defaults to 5, it does not specify a maximum, or whether or not the size of each chunk is derived from the total filesize / concurrency.
I trolled the source code and the comment is pretty much the same as the docs:
Set the concurrency level to use when uploading parts. This affects
how many parts are uploaded in parallel. You must use a local file as
your data source when using a concurrency greater than 1
So my functional build looks like this (the vars are defined by the way, this is just condensed for example):
use Aws\Common\Exception\MultipartUploadException;
use Aws\S3\Model\MultipartUpload\UploadBuilder;
$uploader = UploadBuilder::newInstance()
->setClient($client)
->setSource($file)
->setBucket($bucket)
->setKey($file)
->setConcurrency(30)
->setOption('CacheControl', 'max-age=3600')
->build();
Works great except a 200mb file takes 9 minutes to upload... with 30 concurrent connections? Seems suspicious to me, so I upped concurrency to 100 and the upload time was 8.5 minutes. Such a small difference could just be connection and not code.
So my question is whether or not there's a concurrency maximum, what it is, and if you can specify the size of the chunks or if chunk size is automatically calculated. My goal is to try to get a 500mb file to transfer to AWS s3 within 5 minutes, however I have to optimize that if possible.
Looking through the source code, it looks like 10,000 is the maximum concurrent connections. There is no automatic calculations of chunk sizes based on concurrent connections but you could set those yourself if needed for whatever reason.
I set the chunk size to 10 megs, 20 concurrent connections and it seems to work fine. On a real server I got a 100 meg file to transfer in 23 seconds. Much better than the 3 1/2 to 4 minute it was getting in the dev environments. Interesting, but thems the stats, should anyone else come across this same issue.
This is what my builder ended up being:
$uploader = UploadBuilder::newInstance()
->setClient($client)
->setSource($file)
->setBucket($bicket)
->setKey($file)
->setConcurrency(20)
->setMinPartSize(10485760)
->setOption('CacheControl', 'max-age=3600')
->build();
I may need to up that max cache but as of yet this works acceptably. The key was moving the processor code to the server and not relying on the weakness of my dev environments, no matter how powerful the machine is or high class the internet connection is.
We can abort the process during upload and can halt all the operations and abort the upload at any instance. We can set Concurrency and minimum part size.
$uploader = UploadBuilder::newInstance()
->setClient($client)
->setSource('/path/to/large/file.mov')
->setBucket('mybucket')
->setKey('my-object-key')
->setConcurrency(3)
->setMinPartSize(10485760)
->setOption('CacheControl', 'max-age=3600')
->build();
try {
$uploader->upload();
echo "Upload complete.\n";
} catch (MultipartUploadException $e) {
$uploader->abort();
echo "Upload failed.\n";
}
(ACF9)
Unless there's an option I'm missing, the "Log Slow Pages Taking Longer Than [n] Seconds" setting isn't useful for front-controller based sites (e.g., Model-Glue, FW/1, Fusebox, Mach-II, etc.).
For instance, in a Mura/Framework-One site, I just end up with:
"Warning","jrpp-186","04/25/13","15:26:36",,"Thread: jrpp-186, processing template: /home/mysite/public_html_cms/wwwroot/index.cfm, completed in 11 seconds, exceeding the 10 second warning limit"
"Warning","jrpp-196","04/25/13","15:27:11",,"Thread: jrpp-196, processing template: /home/mysite/public_html_cms/wwwroot/index.cfm, completed in 59 seconds, exceeding the 10 second warning limit"
"Warning","jrpp-214","04/25/13","15:28:56",,"Thread: jrpp-214, processing template: /home/mysite/public_html_cms/wwwroot/index.cfm, completed in 32 seconds, exceeding the 10 second warning limit"
"Warning","jrpp-134","04/25/13","15:31:53",,"Thread: jrpp-134, processing template: /home/mysite/public_html_cms/wwwroot/index.cfm, completed in 11 seconds, exceeding the 10 second warning limit"
Is there some way to get query string or post details in there, or is there another way to get what I'm after?
You can easily add some logging to your application for any requests that take longer than 10 seconds.
In onRequestStart():
request.startTime = getTickCount();
In onRequestEnd():
request.endTime = getTickCount();
if (request.endTime - request.startTime > 10000){
writeLog(cgi.QUERY_STRING);
}
If you're writing a Mach-II, FW/1 or ColdBox application, it's trivial to write a "plugin" that runs on every request which captures the URL or FORM variables passed in the request and stores that in a simple database table or log file. (You can even capture session.userID or IP address or whatever you may need.) If you're capturing to a database table, you'll probably not want any indexes to optimize for performance and you'll need to rotate that table so you're not trying to do high-speed inserts on a table with tens of millions of rows.
In Mach-II, you'd write a plugin.
In FW/1, you'd put a call to a controller which handles this into setupRequest() in your application.cfc.
In ColdBox, you'd write an interceptor.
The idea is that the log just tells you what pages arw xonsostently slow sp ypu can do your own performance tuning.
Turn on debugging for further details for a start.