Gentoo Archives: gentoo-user

From: Robert David <robert.david@××××××.net>
To: gentoo-user@l.g.o
Cc: antlists <antlists@××××××××××××.uk>
Subject: Re: [gentoo-user] [OT] Using an odd number of drives in ZFS RaidZ
Date: Sun, 04 Jul 2021 10:56:19
Message-Id: 8267067.vXUDI8C0e8@robert-notebook
In Reply to: Re: [gentoo-user] [OT] Using an odd number of drives in ZFS RaidZ by antlists
1 On Thursday, July 1, 2021 5:01:25 PM CEST antlists wrote:
2 > On 01/07/2021 14:47, Robert David wrote:
3 > > Hi Frank,
4 > >
5 > >
6 > >
7 > > In any of my data arrays I have long time migrated off the RAIDZ to the
8 > > MIRROR or RAID10. You will find finally that the RAIDZ is slow and not
9 > > very flexible. Only think you gain is the extra space in constrained
10 > > array spaces. For RAID10 it is much easier to raise the size, just
11 > > resilvering to new bigger disks, removing old and expanding. The
12 > > resilvering speed is magnitude faster.
13 > >
14 > > And anyway much easier to recover
15 > > in cases of failure.
16 >
17 > ARE YOU SURE???
18 >
19 > The standard mirror does not cope with corruption very well. Lose a disk
20 > and resilvering is fast. Corrupt the data, and you'll be tearing your
21 > hair out why things go wrong randomly, with no automated way, even once
22 > you've realised what's happened, to recover your data other than a
23 > restore from backup.
24
25 Yes I'm sure. What I meant easier is that the pool is much easier to
26 handle and recover.
27
28 For RAIDZ1 the thing you mention for MIRROR is the same, only it is
29 multiplied with the amount of disks. So if 1 disk fail and you resilver,
30 then all the remaining disks spinning to populate the spare. If any of
31 them fails, then you are screwed. RAIDZ2 is better in this space and in
32 case of 4 disks it is better when it comes to resiliency (for 10 disks
33 it may not be true), but you lose the flexibility.
34
35 Also time to resilver under RAIDZ is much slower, which means longer
36 time under unprotected pool. It is always needed to decide what workload
37 you are serving and how precious the data are.
38
39 For data like movies RAIDZ1 is enough I think.
40
41 Also it is good to check the SMART data time to time to see the amount
42 of error corrections (some are ok, but highly rising no). Solaris has
43 FMA for this to kick in spare. Under home environment it is fine to
44 check it time to time and consider new disk before the old one
45 completely dies. This reminds me I need to buy new disk to my home NAS :)
46 (because of the rising corrections).
47
48 And finally, always do backups for the data you are about to save. I got
49 raspberry pi with attached USB JBOD with two disks serving as backup
50 station. It is not fast to be a real NAS, but to do send/receive of
51 incremental snapshots it is enough. I automaticly sync there the
52 datasets that are worth not to lose (photos, documents, etc), many times
53 these datasets are also the ones that are not such big.
54 Ideally put this backup station to some remote location (or at least
55 different room).
56
57 Robert.
58
59 >
60 > > If you really need the additional space, consider adding second jbod
61 > > with another disks.
62 >
63 > That'd be my approach - migrate a load of stuff off onto another disk
64 > elsewhere, but that's not what the OP wants to do.
65 >
66 > > Robert.
67 >
68 > Cheers,
69 > Wol