Network Management

So pretty beefy server when nighltly maintenance runs or a manual backup is taken, the disk utilization graph is 100% or near 100%.

Is that normal?

Pasquale

yep, pretty normal. better metrics to look at are the iowait and also the duration of the nightly backup. High amounts of iowait are indicative of your disk not keeping up, i.e. not enough iops.

regards

-jeff

Hey Jeff,

I’ll take a look, the backup starts at midnight + 1 min. The last backup finished at 1pm that same day (yesterday)
This AMP is monitoring a large public hotspot with over 1000 locations so lots of users in and out, rrds, etc.. My HDR is set pretty low I would say.
The backups have been growing about 300Mb each night. Now at 1.8Gb.

Which graph would you be looking at for IO wait.

I’ve got a RAID-10 (perc H730p 2Gb) with 8 x 300Gb 15k SATA drives and 128Gb of RAM.
2 x Intel(R) Xeon(R) CPU E5-2630 v3 @ 2.40GHz Hyper-Threaded 8 Cores 20480 KB cache.

I am hoping performance is not an issue.

I had a hell of a time installing Airwave on it.
Ran into the same issue here: http://community.arubanetworks.com/t5/Unified-Wired-Wireless-Access/Airwave-Install-no-usable-disks-found/m-p/239316/highlight/true#M50264

you do need to a bit careful with hotspots vs. Airwave, especially around data retention. The RRD directory can become very large as the number of unique clients increase within the retention intervals. Try to keep your retentions as low as (practically) possible. A customer of mine with 1500 devices has around 500 GB of files under rrd, for around 2 million unique client macs (per Airwave) - with only 60 day retention. The database dump and the tar of the rrd directory (as part of the nightly maint) are the two operations that really take a long time.

 

Check the system cpu util in the performance section - iowait is shown in yellow there. You should expect some of it during nightly maint, but you want it to be minimised.

 

regards

-jeff

 

Oh I know. I've got previous experience with Airwave and hotspots where Dan C helped me out quite a bit.

My HDR is 31 for client data retention and 31 for inactive and active client association and user data so not high. Everything else is 14 or lower.

My CPU I/O wait is fairly low. Avg 3.5%, idle is 93.2%.

I had never noticed the disk utilization almost maxing out during nightly maintenance so that's what got me worried.

We did end up upgrading the raid controller firmware which improved disk read/writes quite a bit but because we only have 8 monitoring ~1040 devices with an average of 8K to 12K clients a day, its doing what it can.

I have about 6735766 rrd files on the system and it has been up for more than 2 weeks or so.
RRD file size varies between ~416k to ~488k which amounts to about 390Gb of rrd files.

I'll be keeping an eye on disk space as well.

Thanks jeff for you input as always.