Separate names with a comma.
Discussion in 'Tutorials and Guides' started by drmike, Jun 26, 2013.
Oh dear. The thought of letting a node get that low makes me cringe.
Monitoring diskspace is often hard, I mostly use PRTG by Paessler and their default settings are warning at 25% and error at 10%.
10% of a 18TB array is just wrong (in the other way).
In my case this is 4x 600GB of SAS 15k RPM in RAID10 array. A bit more than 1TB of usable disk space. 10GB may be little low, but still enough to live-migrate to another Node in the VLAN if you have this script in hourly cron.
Well, you can always put it to 100 GB or whatever.
We use munin to warn for less than 10% left. It never happens for node storage, but for the root partition when logging goes out of control.
That is with free VPSs. No overselling (not even bandwidth)
Node specs: 32gb ram, E3-1230, 4x1TB with mdadm RAID10, $159.99/mo
I wonder how you make back the $160, advertising?
I think he makes even more, must pay his bills, I presume...
It's KVM though so the disk space is assigned on LV's. Yes you can oversell it but that's just asking for a FSCK. <_<
Yes. I could easily have another paid for by ads if I felt like it was needed.