PDE Prep: Cloud Managed Apache Spark Cluster Operations and Maintenance Reviews

Vinicius H. · Reviewed almost 5 years ago

Luciana B. · Reviewed almost 5 years ago

Bruno C. · Reviewed almost 5 years ago

Jorge S. · Reviewed almost 5 years ago

Bruno C. · Reviewed almost 5 years ago

Couldn't complete task 4 because versions of cluster are different, there is no 1.3 version

Luis Arinobu O. · Reviewed almost 5 years ago

mjtelco-test-2のjobでエラーになりませんでした。dataprocとしては処理性能が向上したので良いとは思いますが、シナリオの見直しが必要と感じました。

秀吾 沼. · Reviewed almost 5 years ago

Again for the 3rd time, we try to execute the lab. A mistake should be raised when creating and running job mjtelco-test-2, anyway even following the relevant instructions the job,expected to fail for lack of memory, runs perfectly and the result is not recorded in the score.

Nicola P. · Reviewed almost 5 years ago

Did the lab again to try again. Job does not fail where it should fail with available images. gcloud dataproc clusters create mjtelco \ --project qwiklabs-gcp-01-15432c8f587b \ --region=us-central1 \ --zone=us-central1-a \ --master-machine-type=n1-standard-2 \ --worker-machine-type=n1-standard-2 \ --image-version=1.4-debian10 \ gcloud dataproc jobs submit \ pyspark gs://qwiklabs-gcp-01-15432c8f587b/benchmark.py \ --id=mjtelco-test-2 \ --cluster=mjtelco \ --max-failures-per-hour=1 \ --region=us-central1 \ -- 220 After submit, job executes perfectly, no java out of memory error here.

Dirk H. · Reviewed almost 5 years ago

The lab is become old and has multiple issues 1. The image used in the lab is no longer listed in GCP, so I had to take a higher version image. 2. The first job ran fine, the second job also, even though the dataproc cluster was created with the provided settings. 3. Since the second job ran succesful, the objective could not be completed, re-running the job was not possible due to already having ran a job with the same job-id, other naming would mean the lab couldn't validate the success of the objective. Overall the idea is good, but this lab is needs to be updated.

Dirk H. · Reviewed almost 5 years ago

The step with "failed job" has a problem: using 220 as argument doesn't make the job fail anymore.

Daniel S. · Reviewed almost 5 years ago

Md I. · Reviewed almost 5 years ago

Md I. · Reviewed almost 5 years ago

Thauren H. · Reviewed almost 5 years ago

Finished. Good Job

aulia s. · Reviewed almost 5 years ago

Finished

aulia s. · Reviewed almost 5 years ago

Lab is broken and outdated. There is no more version 1.3 image and copying just the .py file is not enough - it requires the entire .* of the directory. This was good enough for all tasks, except for the "demonstrate failure" one - the way GCP works now does not cause OOM Errors anymore (even when I tried getting the number of partitions to extreme - it tells me about 22TiB memory and maxes on 12 GB allocated, but still keeps on running chunk by chunk.) Trying to override driver setting didn't help either and, since the lab itself was expecting a very specific failure, neither did failing the job differently (by referring it to a non-existing file.)

Aleksandr Z. · Reviewed almost 5 years ago

Arthur M. · Reviewed almost 5 years ago

Karthick K. · Reviewed almost 5 years ago

super

Yogee P. · Reviewed almost 5 years ago

LEUNG, C. · Reviewed almost 5 years ago

très bien fait

Koniba J. · Reviewed almost 5 years ago

one of the tasks took about 45 minutes to fail which is waste of time

翔中 張. · Reviewed almost 5 years ago

sibatyu9793 s. · Reviewed almost 5 years ago

Demetrius T. · Reviewed almost 5 years ago