Hello,
Reading documentation regarding setting up offloading MySQL/MariaDB to a stand alone server and noted the option to implement "jumbo frames". Are there any clearer specifications for when jumbo frames should be introduced? Certain number of hosts/services or certain number of checks per second or average check per second and latency thresholds which should be key indicators etc. Obviously, if our environment doesn't require the extra overhead/operational work to set up on jumbo frames in network we'd prefer not to do it
Thanks so much Nagios team,
Deirdre
You really only need jumbo frames if you're offloading the DB, because of the communication between the XI server and the DB machine.
If you keep the DB local to the XI server, it's not an issue. And we have plenty of customers that run quite large installations with a local DB, also.
--Jeffrey
As of May 25th, 2018, all communications with Nagios Enterprises and its employees are covered under our new Privacy Policy.
Be sure to check out our Knowledgebase for helpful articles and solutions!
Hi Jeffrey,
We are planning to offload our database as we'll be soon exceeding 1000 hosts and well over 10000 services. We are making this decision based on the Nagios XI documentation which states:
Nagios XI Hardware System Requirements / Database Requirements not specified explicitly
More than 500 hosts, more than 2,500 services:At least 120 GB disk space, more than 4 cores, and more than 8 GB RAM.
If monitoring over 1000 hosts or 5000 services, a dedicated physical server is recommended.
Looking for the standards/threshold metrics from Nagios on when this jumbo frame option is most appropriate.
My experience with this is that once you get close to 20K total checks you'll likely need to move the DB back local because the additional network latency causes slower kernel message queue processing (if running the old NDO2DB). Even with Jumbo Frames enabled it doesn't seem to have a huge impact on the speed of processing the kernel message queue according to another technician who tested that setup out.
What is the main reason you're offloading it? Are you seeing any issues that prompted you to want to offload it? If so, what?
We aren't experiencing any issues to date but are planning our 2021 right sizing roadmap strategy now so we can pro-actively size the environment vs. being reactive. We expect in 2021 to exceed 5k hosts and 12000k services by the time we migrate some legacy monitoring over.
Thank you for your feedback / information re: kernel latency. Something for us to add to our considerations as we determine our direction.
Hi there,
Pro-actively isntalling RAM disk as we speak. Already have it running on our DEV environment and testing now. Will be setting this up on PRD box end of the month.
You can go ahead and close this post as we have another of mine that has ended up overlapping here. Thanks so much for all of your help.
danderson6@unum.com wrote:Hi there,
Pro-actively isntalling RAM disk as we speak. Already have it running on our DEV environment and testing now. Will be setting this up on PRD box end of the month.
You can go ahead and close this post as we have another of mine that has ended up overlapping here. Thanks so much for all of your help.
