unlucky1 wrote: it seems to fail at around 8000ish.
I suppose that's 8192. (2^13). When only the number of processes is the limit, and not the available memory, you could increase this number:
- Code: Select all
echo 32768 >/proc/sys/kernel/pid_max
unlucky1 wrote: I could also change out 2 of the disks every week or so and keep a rotating back-up for disaster recovery.
That is not a good idea. Nor the connectors on the disks, neither the connectors in the Lacie are designed for that. But you could use the e-sata connectors for that.
unlucky1 wrote:The truth is I really don't know what's best
"Best". That's a complicated word. I could tell you what's best, but you would react: "What! 1 million euro's
a week?"
My personal preference against backups is to automate erverything, as I'm just too unorganized to rely on daily/weekly human initiated actions, and to have at least one offsite backup. And I'm not a fan of raid. Raid is overrated.
So far I haven't had the need to create big volumes (bigger than a disk), but if I had, I would have used mhddfs or something similar, and not raid.
I *think* your best option is a 2 disk mhddfs volume to contain the data, a 2 or 3 disks volume containing rsnapshot backups (And we should investigate if "the big rsnapshot trick"® (usage of hardlinks, so that each individual file is only one time present in numerous complete backups) works on a mhddfs volume.)
Further an offsite NAS which gets a weekly backup over the internet. Of course you need some bandwidth for it. How much depends on how changes in your data set can be expected in a week.
This can not be done on Lacie firmware. It can be done on fvdw's firmware, if he is willing to spend some time on it.
And it can be done on any Linux distro which could run on your box, like Debian, Arch, Ubuntu, ... But for all those cases first some serious time should be invested to get that distro running on the Lacie.