Issues with Introscope Collector in AWS

Document ID : KB000112011
Last Modified Date : 28/08/2018
Show Technical Document Details
Question:
The APM environment is 10.5.1. The Collector (running on RHEL 7.0) that we have installed in AWS is having a massive amount of I/O which is causing it to exceed the EFS Burst Credit Balance.  This causes throttling of the disk I/O on that Collector which brings the whole APM environment to a crawl.  I'm working with our DevOps team that supports the Collector.  Here are some notes that the DevOps engineer provided.

Notes:
We have an EM that has been dockerized and deployed on AWS using AWS ECS (container orchestration), EC2 (virtual machines), and EFS (basically a NFS).  Our docker deployment currently has the 2 folders mounted to EFS:
/opt/monitoring/installed/collector/data
/opt/monitoring/installed/collector/traces

We also have logs captured from the “${INSTALLDIR}/Introscope_Enterprise_Manager” being loaded to log aggregation, so not sure if we need the traces folder to be mapped to EFS or not.  Please confirm or not.

The issue we see is extreme IO on EFS (NFS) which results in us using up all of our credit balance, which results in throttling of IO, and causes a lot of performance problems at the MOM.  I don’t believe we have a very high load on the APM, yet we are going through ~150-200gb of IO a day.  Is this expected?
Answer:
https://docops.ca.com/ca-apm/10-5/en/ca-apm-sizing-and-performance/apm-sizing-and-performance-on-amazon-web-services

According to the docs, APM is certified to be used on EBS and not EFS. EFS is equivalent to NFS and not at all recommended for smartstor. See the section "SmartStor on Network File Systems" in 

https://docops.ca.com/ca-apm/10-5/en/ca-apm-sizing-and-performance/system-information-and-requirements/ca-apm-data-storage-requirements#CAAPMDataStorageRequirements-EnterpriseManagerInternalDatabaseSettingsandCapacity