2019-10-10 07:07:28
why not three mirrors? quicker to recover
Rotonen
2019-10-10 14:44:45
mdadm doesn't support arbitrary reed-solomon, I think
Fireduck
2019-10-10 14:55:18
well, if disks die of old age eventually, good luck recovering the raid before more fail
Rotonen
2019-10-10 15:32:02
yeah, it isn't stuff I actually need
Fireduck
2019-10-10 15:32:21
things I actually need fit in much smaller storage
Fireduck
2019-10-10 15:32:56
I never trust a single filesystem with anything I actually need
Fireduck
2019-10-10 15:35:40
yep, my rule is to have the backups on different tech
Rotonen
2019-10-10 16:10:37
yeah, my backups are rsync to a freebsd zfs box which does daily and monthly snapshots
Fireduck
2019-10-10 16:10:45
the monthly snapshots are streamed off to S3
Fireduck
2019-10-10 16:11:03
Then I have another freebsd ec2 system that loads the snapshots from S3
Fireduck
2019-10-10 16:11:16
to ensure that I didn't screw it up and the snapshot streams actually load
Fireduck
2019-10-10 16:11:30
and also, it means that I can access the files quickly if my house is lost
Fireduck
2019-10-10 16:11:54
I have snapshots that go back to 2006, so it takes a while to load
Fireduck
2019-10-10 16:12:42
Although, to be fair the first year or so of snapshots where copied in from some other system. Haven't been using zfs quite that long.
Fireduck
2019-10-10 17:28:26
ever considered glacier?
Rotonen
2019-10-10 17:29:02
yeah, I am using that (via S3)
Fireduck
2019-10-10 17:29:15
the super cheap deep one
Fireduck
2019-10-10 17:29:23
it will hurt if I ever need to restore from that
Fireduck
2019-10-10 17:30:08
i’ve never heard of anyone having done disaster recovery off it
Rotonen
2019-10-10 17:30:19
curious for war stories
Rotonen
2019-10-10 18:42:10
I was on glacier a number of years ago and had to do a restore
Fireduck
2019-10-10 18:42:22
it was as predicted, slow to start but then went fine
Fireduck
2019-10-10 18:42:32
well, there was a shockingly high bill
Fireduck
2019-10-10 18:42:58
their billing was basically insane, I had to read it many times before I understood.
Fireduck
2019-10-10 18:43:31
Basically, it was if you pulled a file of size X, they would assume the entire file transferred in some 5 minute window and that would result in a bandwidth number
Fireduck
2019-10-10 18:43:51
then you were billed for that bandwidth number for like 24 hours or something non-sensensical
Fireduck
2019-10-10 18:44:08
they have made that much simpler now, just charging based on GB restored
Fireduck
2019-10-10 18:44:58
I do wonder what sort of redundancy they can manage at the prices they are charging
Fireduck
2019-10-10 18:45:15
but it is probably a loss leader or nearly so to get people's data into AWS platform
Fireduck
2019-10-10 21:29:55
it is the cheap beer of their supermarket
Rotonen