Installing and using the pcp-zeroconf package for Performance Co-Pilot (PCP)

Installation

• pcp-zeroconf is intended to simplify the installation and configuration of the most commonly needed PCP features in customer support environments, where we need comprehensive long term logging of system level performance data with minimal overheads. The pcp-zeroconf RPM package installs a minimal set of dependencies for the server side data collection and does not include client tools. The package is part of PCP from version 3.11.10 onwards, and is also included in RHEL 7.4 and later, as well as RHEL 8. There is currently no official RHEL6 package.
• If the system is not registered/subscribed please see: How to register and subscribe a system to the Red Hat Customer Portal using Red Hat Subscription-Manager
• For RHEL 7 and 8, pcp-zeroconf is in the base repository and no additional repository modifications are needed:

# yum install pcp-zeroconf

• As the name suggests, that's it - there are no further installation or configurations steps required in order to collect performance data archives. It is important to make sure there is sufficient space on the /var filesystem (or root filesystem if /var is not a submount) for on-going log collection - approximate 5GB will be required on most small servers. On large servers, at least 10GB of space on /var will be used on an on-going basis. By default, PCP archive logs are automatically culled after 14 days - see pmlogger_daily(1) for instructions on changing this if needed.

Services and configuration

• The post-install scripts enable the pmcd, pmlogger and pmie services under the control of systemd. The performance data are captured below /var/log/pcp and the performance data logs are automatically rolled-over and compressed on a daily basis. If and when required, these logs can be tarred up and attached to a Red Hat customer support case for analysis by Red Hat support engineers.

• The performance data archive logs captured by pcp-zeroconf include almost all of the Linux kernel metrics for CPU usage, network, filesystem, disk and memory subsystems. In addition, the optional nfs-client, XFS per-device and proc (per-process) PCP agents are enabled by the post-install script and logged by the pmlogger service.

• The captured logs contain sufficient performance metrics to be used by most of the PCP monitoring tools (see references below).

Note

One can also install the Performance Co-Pilot by simply installing the 'pcp' rpm package but please note the below mentioned differences among these two packages:

• 'pcp-zeroconf' and 'pcp' package follows different logging interval - 10 sec & 1 min respectively.
• 'pcp-zeroconf' by default enables the process level logging whereas with the 'pcp' package one has to enable it manually.
• As the name suggests 'pcp-zeroconf' by default enables the 'pmcd' and 'pmlogger' services whereas with the 'pcp' one has to enable them manually.
• 'pcp-zeroconf' pulls 'pcp' package as a dependency and make the above listed changes in the overall configuration of Performance Co-Pilot.

So on RHEL systems it is always recommended to install the 'pcp-zeroconf' package instead of 'pcp'.

References

• Index of Performance Co-Pilot (PCP) articles, solutions, tutorials and white papers
• How do I gather performance data logs to upload to my support case using Performance Co-Pilot (PCP)
• Side-by-side comparison of PCP tools with legacy tools
• How to use a non-default PMDA with PCP?
• How can I run PCP services in a Docker container?

White Papers, Guide Books, Case Studies and Presentations

• Introduction to storage performance analysis with PCP
• Performance Co-Pilot Quick Reference Guide
• Performance Co-Pilot User's and Administrator's Guide
• Performance Co-Pilot Programming Guide
• Performance Co-Pilot Tutorials and Case Studies
• Performance Co-Pilot + a RHEL Case Study