Splunk Enterprise

Splunk Forwarder Performance

smahtha
Engager

We have a "complicated" directory structure for logging. I wanted to understand the impact on performance for Splunk Forwarder for the below case.

Lets say our logs have the below structure /foo/bar*/*/abc/def/ghi*.log and /foo/bar*/*/abc/def/jkl*.log

Now if we use the above wildcard structure for monitoring purpose, Splunk will translate it as
[monitor://foo/]
whitelist: <appropriately set for ghi*.log and jkl*.log files]

This will effectively monitor the directory and all the subdirectories inside /foo/. And as per my understanding, whenever a new files is added or updated anywhere inside /foo/ it will be compared against the whitelist by Splunk forwarder to decide whether to index it or not.

The number of files eligible for logging is not a lot (hardly 3-4). But the number of files that can be created inside /foo/ (all of which are eligible for whitelist comparison) can be in thousands.

Can anyone give a rough estimate of [if understood and known] the impact of this kind of directory structure will have on Splunk forwarder performance.

Tags (1)
0 Karma
1 Solution

smahtha
Engager

We did a performance profling to understand the impact of the "directory structure and number of files eligible for whitelist comparison" on Splunk forwarder performance. The overall idea is this:
a) If a higher level directory is monitored instead of 'n' lower level directories [eg. /foo/ vs (/foo/bar1/123/abc/def/ and /foo/bar2/123/abc/def/ ... 10-times ... /foo/bar10/123/abc/def/)] CPU usage of the forwarder is higher.
b) If the number of files eligible for whitelist comparison is more then also CPU usage is greater.

Now some hard numbers.
We compared monitoring /foo/ vs ~50 /foo/bar1/123/abc/def/. In the case of /foo/ the number of files eligible for monitoring were around ~1000 whereas the number of files eligible for monitoring in the case of 50 /foo/bar1/123/abc/def/ was ~200.

The CPU usage in the first case rose up to around 10% whereas in the second case it was around 2%. The memory footprint remained roughly the same arounf 120MB.

View solution in original post

smahtha
Engager

We did a performance profling to understand the impact of the "directory structure and number of files eligible for whitelist comparison" on Splunk forwarder performance. The overall idea is this:
a) If a higher level directory is monitored instead of 'n' lower level directories [eg. /foo/ vs (/foo/bar1/123/abc/def/ and /foo/bar2/123/abc/def/ ... 10-times ... /foo/bar10/123/abc/def/)] CPU usage of the forwarder is higher.
b) If the number of files eligible for whitelist comparison is more then also CPU usage is greater.

Now some hard numbers.
We compared monitoring /foo/ vs ~50 /foo/bar1/123/abc/def/. In the case of /foo/ the number of files eligible for monitoring were around ~1000 whereas the number of files eligible for monitoring in the case of 50 /foo/bar1/123/abc/def/ was ~200.

The CPU usage in the first case rose up to around 10% whereas in the second case it was around 2%. The memory footprint remained roughly the same arounf 120MB.

Got questions? Get answers!

Join the Splunk Community Slack to learn, troubleshoot, and make connections with fellow Splunk practitioners in real time!

Meet up IRL or virtually!

Join Splunk User Groups to connect and learn in-person by region or remotely by topic or industry.

Get Updates on the Splunk Community!

Event Series: Splunk Observability Metrics Cost Optimization

Balancing Scale and Spend: Gaining Control Over High-Volume Metrics in Splunk Observability Cloud As ...

Kick the Tires Before You Commit: A Hands-On Tour of the Splunk Observability Cloud ...

Evaluating an enterprise observability platform usually goes like this: fill out a form, get a free trial with ...

Deep insights, no barriers: Splunk Observability Cloud Free Edition

As software delivery cycles continue to accelerate, observability shouldn’t be a luxury — it should be a ...