| View previous topic :: View next topic |
| Author |
Message |
bunder
Bodhisattva
Joined: 10 Apr 2004
Posts: 5947
|
 Posted: Sat May 02, 2009 6:01 am Post subject: convert stripe to raid5? (i think i finally got it) |
 |
|
anyone know if it's possible to convert a two disk stripe into a four disk raid5? all disks are the same make/model, done with mdadm.
i'm guessing "not without losing your data", but i'm hoping to retain the data, and not have to buy a 500gb drive just to make a temporary backup.
thanks
_________________
| Neddyseagoon wrote: | | The problem with leaving is that you can only do it once and it reduces your influence. |
banned from #gentoo since sept 2017
Last edited by bunder on Thu May 07, 2009 12:21 am; edited 3 times in total |
|
| Back to top |
|
 |
NeddySeagoon
Administrator
Joined: 05 Jul 2003
Posts: 54805
Location: 56N 3W
|
| Posted: Sat May 02, 2009 10:48 am Post subject: |
|
|
bunder,
I assume you mean a two disc raid0 to a four disk raid5 ?
Not directly. You can make the raid5 in degraded mode, which only needs 3 drives but you can't take one out the the raid0 set directly.
If we are talking kernel raid0 and its less than half full, you may be able to do some data juggling to go from raid0 to single drive and put the freed drive into the raid5 set.
_________________
Regards,
NeddySeagoon
Computer users fall into two groups:-
those that do backups
those that have never had a hard drive fail. |
|
| Back to top |
|
|
Cyker
Veteran
Joined: 15 Jun 2006
Posts: 1746
|
| Posted: Sat May 02, 2009 10:59 am Post subject: Re: convert stripe to raid5? |
|
|
| bunder wrote: | anyone know if it's possible to convert a two disk stripe into a four disk raid5? all disks are the same make/model, done with mdadm.
i'm guessing "not without losing your data", but i'm hoping to retain the data, and not have to buy a 500gb drive just to make a temporary backup.
thanks |
Theoretically you can, but not the way I suspect you mean (i.e. converting the RAID0 to RAID5 then expanding it onto the 2 extra drives).
The way to do it the way I know of would go something like this: (Disclaimer; I haven't tried this!)
First you need to clear out the 'new' two drives and create a RAID5 array on them with something like:
| Code: | | mdadm --create /dev/md1 --level=5 -n 2 /dev/hdc1 /dev/hdd1 |
Obviously, this is the time to add things like --chunk=128 or whatever if you want to tweak the underlying properties of the RAID array.
Once that's done, hopefully you will have a degraded RAID5 array that you can you can format the partition or LVM it or whatever, e.g. ...
| Code: | | mke2fs -j -O dir_index -L cyServRAID5 -v -b 4096 -E stride=16 /dev/md1 |
(Obviously tweak the settings here; If you used --chunk=128, you'll want --stride=32 here, and if you change the -b blocksize, then you need to calculate the stride yourself. Or use whatever is appropriate for xfs/reiser or lvm etc.!)
...and copy your RAID0 to; Do that - mount it, copy over, TEST THAT YOU CAN READ IT BACK.
--- Start of Danger Zone!  ---
Then you need to zap the existing RAID0:
| Code: | | mdadm --stop /dev/md0 |
Clear it out from the mdadm.conf if you are using it, then set up the freed disks as you normally would prepare them for addition to an array, if needed.
From there it's the standard --add and very lengthy and nail-biting --grow command to expand the array:
| Code: | mdadm /dev/md1 --add /dev/hdc1
mdadm --grow /dev/md1 -n=3 --backup-file=/tmp/mdadm.grow.backup -z max
mdadm /dev/md1 --add /dev/hdd1
mdadm --grow /dev/md1 -n=4 --backup-file=/tmp/mdadm.grow.backup -z max |
Or, if you feel brave, this might work too:
| Code: | mdadm /dev/md1 --add /dev/hdc1 /dev/hdd1
mdadm --grow /dev/md1 -n=4 --backup-file=/tmp/mdadm.grow.backup -z max |
NB: Either way, the --backup-file MUST point to a file that isn't on the array, so e.g. if your /tmp is on the array, you will need to specify a different place.
After that finishes (It will take hours, possibly days! ), you just need to resize the FS to fit into the new space.
With ext3, it is easy:
| Code: | | resize2fs -p /dev/md1 |
The -p just displays a completion status; ith no other args it just defaults to expanding the FS to the full size of the partition.
If you have anything like reiser or xfs, or some exotic LVM thing then you'll have to read up on it yourself, but I'm sure it's not much harder 
Normally I preach about making backups, but I figure if you could you wouldn't be asking how to do it this way...  |
|
| Back to top |
|
|
bunder
Bodhisattva
Joined: 10 Apr 2004
Posts: 5947
|
| Posted: Sat May 02, 2009 12:12 pm Post subject: Re: convert stripe to raid5? |
|
|
| Cyker wrote: | | Theoretically you can, but not the way I suspect you mean (i.e. converting the RAID0 to RAID5 then expanding it onto the 2 extra drives). |
actually, the way i was thinking of doing it was convert the raid 0 to a degraded raid 4 (no parity disk), add the two disks and then convert it to raid 5... but now that i know you can make a raid5 with two disks, i might be able to do that instead, skipping a step.
my current config looks like this:
| Quote: | DEVICE /dev/hdc1 /dev/hdd1
DEVICE /dev/hdc2 /dev/hdd2
ARRAY /dev/md0 level=raid0 num-devices=2 devices=/dev/hdc1,/dev/hdd1
ARRAY /dev/md1 level=raid0 num-devices=2 devices=/dev/hdc2,/dev/hdd2
|
all four drives are WD2500JB's and i only plan on retaining one partition on the current array, which contains 395gb of mythtv storage...
dunno, i'm tired and don't know how to finish this post...  yeah.
_________________
| Neddyseagoon wrote: | | The problem with leaving is that you can only do it once and it reduces your influence. |
banned from #gentoo since sept 2017 |
|
| Back to top |
|
|
Cyker
Veteran
Joined: 15 Jun 2006
Posts: 1746
|
| Posted: Sat May 02, 2009 12:35 pm Post subject: |
|
|
Huh, 4 drives? I only count 2 there??
The problem with what you want to do is that I don't think mdadm can do direct RAID level conversions (Can it??).
For the --grow mode, it specifically says that the --level param is not supported in the man page... |
|
| Back to top |
|
|
bunder
Bodhisattva
Joined: 10 Apr 2004
Posts: 5947
|
| Posted: Sat May 02, 2009 12:39 pm Post subject: |
|
|
| Cyker wrote: | Huh, 4 drives? I only count 2 there??
The problem with what you want to do is that I don't think mdadm can do direct RAID level conversions (Can it??).
For the --grow mode, it specifically says that the --level param is not supported in the man page... |
the other two drives are still in another machine. 
i don't know about conversions between levels and/or grow... that's why i'm here asking i guess.
_________________
| Neddyseagoon wrote: | | The problem with leaving is that you can only do it once and it reduces your influence. |
banned from #gentoo since sept 2017 |
|
| Back to top |
|
|
NeddySeagoon
Administrator
Joined: 05 Jul 2003
Posts: 54805
Location: 56N 3W
|
| Posted: Sat May 02, 2009 1:01 pm Post subject: |
|
|
bunder,
Take the rad0 set(s) and copy the data to a new single drive.
Validate the copy as the next step is destructive. The single drive will become one of the raid5 later.
Using the drives from the raid0 set and the other new drive, make a degraded raid5 set with 3 drives. mdadm will hate you for this as it will detect the existing raid0 set.
Copy the data you moved to the single drive to to the degraded raid5. Validate the copy again as the next step is destructive.
Add the single drive to the raid5 set. When its synched, you have a full strength raid5.
_________________
Regards,
NeddySeagoon
Computer users fall into two groups:-
those that do backups
those that have never had a hard drive fail. |
|
| Back to top |
|
|
Cyker
Veteran
Joined: 15 Jun 2006
Posts: 1746
|
| Posted: Sun May 03, 2009 12:56 am Post subject: |
|
|
| bunder wrote: | | Cyker wrote: | Huh, 4 drives? I only count 2 there??
The problem with what you want to do is that I don't think mdadm can do direct RAID level conversions (Can it??).
For the --grow mode, it specifically says that the --level param is not supported in the man page... |
the other two drives are still in another machine.
i don't know about conversions between levels and/or grow... that's why i'm here asking i guess. |
Ah, okay.
Anyway, I am led to believe that mdadm currently doesn't support changing RAID levels on-the-fly, although there is a possibility it may be added in the future.
Ironically, there is a utility called "raidreconf" which CAN convert between RAID levels but I don't know if you can use it with mdadm RAIDs as it is for the much older raidtab-based arrays.
So right now it looks like your options are:
a) Back up onto a big external 500GB HD and then create the array on the 4 disks (Safest, most expensive if you don't have a 500+GB USB disk), or
b) Create the degraded array etc. in my earlier post. (Costs nothing but lots of time, much more risky)
Neddy's suggestion would be a better and safer bet, but it has the problem that 395GB data won't fit on a 250GB drive |
|
| Back to top |
|
|
bunder
Bodhisattva
Joined: 10 Apr 2004
Posts: 5947
|
| Posted: Sun May 03, 2009 2:00 pm Post subject: |
|
|
| Cyker wrote: | | Neddy's suggestion would be a better and safer bet, but it has the problem that 395GB data won't fit on a 250GB drive |
yeah, neddy and i were talking about that on irc.
but i was thinking about the degraded raid5 on 2 drives though... neddy was saying something about "for every four bytes you lose a bit", and that i should be able to fit the data on the new array... so i might actually do that after all. the machine is on a UPS, so as long as it doesn't go out or the drives don't clunk, i'm good. 
a raid calculator was telling me in the end that 4x210GB=580GB of usable space, so that leaves enough room for now. i can use the other 4x40GB=111GB for storing backups.
still don't know when i want to do it though.
thanks
_________________
| Neddyseagoon wrote: | | The problem with leaving is that you can only do it once and it reduces your influence. |
banned from #gentoo since sept 2017 |
|
| Back to top |
|
|
Cyker
Veteran
Joined: 15 Jun 2006
Posts: 1746
|
| Posted: Sun May 03, 2009 6:52 pm Post subject: |
|
|
Well my thinking was that RAID5 capacity is all drives minus 1, so in your case if you were going for 210GB partitions for each element, it should be about 630GB in total?
(I'm running 4x1TB drives and have about 3TB of usable space).
In RAID5 degraded mode with 2 drives, you should have about 420GB to play with, albeit slowly, and when you add the third drive you'll still have 420GB, but with the redundancy, and then the final drive should push it up to 630GB.
I'd leave it 'til you have a free weekend 'tho! |
|
| Back to top |
|
|
NeddySeagoon
Administrator
Joined: 05 Jul 2003
Posts: 54805
Location: 56N 3W
|
| Posted: Sun May 03, 2009 10:11 pm Post subject: |
|
|
Cyker,
Thats exactly right.
_________________
Regards,
NeddySeagoon
Computer users fall into two groups:-
those that do backups
those that have never had a hard drive fail. |
|
| Back to top |
|
|
bunder
Bodhisattva
Joined: 10 Apr 2004
Posts: 5947
|
| Posted: Sun May 03, 2009 11:58 pm Post subject: |
|
|
damn, why isn't anything on the internet ever right?
as long as i know it'll fit. that's all that matters.
i should have checked this before, but the current partition size is 228gb... i'll have to unload some space and shrink the array, which should be doable, i got a spare 60 and some space on my laptop, so despite taking a while it's still looking doable.
cheers
_________________
| Neddyseagoon wrote: | | The problem with leaving is that you can only do it once and it reduces your influence. |
banned from #gentoo since sept 2017 |
|
| Back to top |
|
|
bunder
Bodhisattva
Joined: 10 Apr 2004
Posts: 5947
|
| Posted: Wed May 06, 2009 11:04 pm Post subject: |
|
|
so after pulling my hair out to find out that mdadm only loaded one of my arrays, i'm stuck here...
| Quote: | prescott ~ # cat /proc/mdstat
Personalities : [raid0] [raid1] [raid6] [raid5] [raid4]
md5 : active raid0 hda2[0] hdb2[1]
29302400 blocks 64k chunks
md4 : active raid0 hda1[0] hdb1[1]
459089280 blocks 64k chunks
md1 : active raid5 hdd2[2] hdc2[0]
39118208 blocks level 5, 64k chunk, algorithm 2 [2/1] [U_]
resync=DELAYED
md0 : active raid5 hdd1[2] hdc1[0]
205077632 blocks level 5, 64k chunk, algorithm 2 [2/1] [U_]
[>....................] recovery = 3.7% (7655708/205077632) finish=86.1min speed=38201K/sec
|
why is it rebuilding? i haven't even put filesystems on the new arrays yet. 
_________________
| Neddyseagoon wrote: | | The problem with leaving is that you can only do it once and it reduces your influence. |
banned from #gentoo since sept 2017 |
|
| Back to top |
|
|
Mad Merlin
Veteran
Joined: 09 May 2005
Posts: 1155
|
| Posted: Wed May 06, 2009 11:34 pm Post subject: |
|
|
| bunder wrote: | so after pulling my hair out to find out that mdadm only loaded one of my arrays, i'm stuck here...
| Quote: | prescott ~ # cat /proc/mdstat
Personalities : [raid0] [raid1] [raid6] [raid5] [raid4]
md5 : active raid0 hda2[0] hdb2[1]
29302400 blocks 64k chunks
md4 : active raid0 hda1[0] hdb1[1]
459089280 blocks 64k chunks
md1 : active raid5 hdd2[2] hdc2[0]
39118208 blocks level 5, 64k chunk, algorithm 2 [2/1] [U_]
resync=DELAYED
md0 : active raid5 hdd1[2] hdc1[0]
205077632 blocks level 5, 64k chunk, algorithm 2 [2/1] [U_]
[>....................] recovery = 3.7% (7655708/205077632) finish=86.1min speed=38201K/sec
|
why is it rebuilding? i haven't even put filesystems on the new arrays yet. |
It needs to initialize the contents on the disks. The same thing happens when you initialize any array with mdadm that has redundancy (RAID 1, 4, 5, 6, and 10).
Edit: You can create a filesystem and start using the array immediately, however.
_________________
Game! - Where the stick is mightier than the sword! |
|
| Back to top |
|
|
bunder
Bodhisattva
Joined: 10 Apr 2004
Posts: 5947
|
| Posted: Wed May 06, 2009 11:38 pm Post subject: |
|
|
| Mad Merlin wrote: | It needs to initialize the contents on the disks. The same thing happens when you initialize any array with mdadm that has redundancy (RAID 1, 4, 5, 6, and 10).
Edit: You can create a filesystem and start using the array immediately, however. |
oh okay, i thought i broke it.
edit: yeah, something looks screwy here...
| Quote: | prescott mnt # df -h
Filesystem Size Used Avail Use% Mounted on
/dev/md4 431G 392G 18G 96% /mnt/md4
/dev/md5 28G 13G 14G 48% /mnt/md5
/dev/md0 193G 188M 183G 1% /mnt/md0
/dev/md1 37G 177M 35G 1% /mnt/md1
|
193G? that's nowhere near the size we mentioned before...
| Quote: | prescott mnt # mdadm --detail /dev/md0
/dev/md0:
Version : 0.90
Creation Time : Wed May 6 19:40:16 2009
Raid Level : raid5
Array Size : 205077632 (195.58 GiB 210.00 GB)
Used Dev Size : 205077632 (195.58 GiB 210.00 GB)
Raid Devices : 2
Total Devices : 2
Preferred Minor : 0
Persistence : Superblock is persistent
Update Time : Wed May 6 19:45:35 2009
State : clean, degraded, recovering
Active Devices : 1
Working Devices : 2
Failed Devices : 0
Spare Devices : 1
Layout : left-symmetric
Chunk Size : 64K
Rebuild Status : 6% complete
UUID : 612a95bf:36ea034c:c6759a90:fd3b495c
Events : 0.6
Number Major Minor RaidDevice State
0 22 1 0 active sync /dev/hdc1
2 22 65 1 spare rebuilding /dev/hdd1
|
_________________
| Neddyseagoon wrote: | | The problem with leaving is that you can only do it once and it reduces your influence. |
banned from #gentoo since sept 2017
Last edited by bunder on Wed May 06, 2009 11:49 pm; edited 1 time in total |
|
| Back to top |
|
|
think4urs11
Bodhisattva
Joined: 25 Jun 2003
Posts: 6659
Location: above the cloud
|
| Posted: Wed May 06, 2009 11:49 pm Post subject: |
|
|
correct me if i'm wrong but you've now a 2disk raid0 and a 2disk raid5 and the later one cannot be easily expanded by additional disks, correct?
Wouldn't it be easier to create the new raid5 in degraded mode already?
the following fully untested...
With that you could add the 2 former raid0 disks to the raid5 array after copying content to it. Something like 'mdadm --create --verbose /dev/md0 --level=5 --raid-devices=4 missing missing /dev/sdc1 /dev/sdd1' should do the trick and when content is copied break raid0 and do 'mdadm /dev/md0 --add /dev/sda1'...
_________________
Nothing is secure / Security is always a trade-off with usability / Do not assume anything / Trust no-one, nothing / Paranoia is your friend / Think for yourself |
|
| Back to top |
|
|
bunder
Bodhisattva
Joined: 10 Apr 2004
Posts: 5947
|
| Posted: Wed May 06, 2009 11:50 pm Post subject: |
|
|
| Think4UrS11 wrote: | | Wouldn't it be easier to create the new raid5 in degraded mode already? |
i thought that's what i was trying to do.
edit: oh wait, i needed to add "missing missing"?
double edit: no, <nevermind i got it>
_________________
| Neddyseagoon wrote: | | The problem with leaving is that you can only do it once and it reduces your influence. |
banned from #gentoo since sept 2017
Last edited by bunder on Wed May 06, 2009 11:58 pm; edited 2 times in total |
|
| Back to top |
|
|
eccerr0r
Watchman
Joined: 01 Jul 2004
Posts: 9882
Location: almost Mile High in the USA
|
| Posted: Wed May 06, 2009 11:52 pm Post subject: |
|
|
Interesting that mdadm supports degenerate 2-disk RAID5's 
Anyway if I'm reading this correctly (2-disk RAID0 to 4-disk RAID5) why couldn't you (or did you? or was this already mentioned?):
1. create an reshapable, new "3"-disk RAID5 array with one missing disk (degraded 3-disk RAID5) using /dev/null or something as the "faulty" disk with the two new disks, and make a new filesystem on it. You should see [UU_] or something like that in mdstat, and your original [UU] for the RAID0. Test that you can actually use the new array, it should work just fine as it waits for another member.
2. copy RAID0 to the degraded RAID5. These two should be the exact same size so it should fit without any trouble, and you need not delete anything.
At this point you have redundant copies of your data. Make sure you diff, as next step is destructive and you'll lose your redundancy temporarily.
3. disassemble the raid 0 and wipe it. Take one of the disks and introduce it to the degraded 3-disk RAID5 as redundancy, and allow it to complete rebuild. You now have a 3-disk redundant RAID5 system [UUU].
4. and finally, obviously, mdadm --grow to reshape the last disk onto the RAID5. Then resize your filesystem.
Done, No additional disks needed, your 2-disk RAID0 is now a 4-disk RAID5. However I would still have that 500G disk handy just in case something goes wrong, but then again you were playing with fire with a RAID0 anyway...
_________________
Intel Core i7 2700K/Radeon R7 250/24GB DDR3/256GB SSD
What am I supposed watching? |
|
| Back to top |
|
|
think4urs11
Bodhisattva
Joined: 25 Jun 2003
Posts: 6659
Location: above the cloud
|
| Posted: Wed May 06, 2009 11:52 pm Post subject: |
|
|
| bunder wrote: | | edit: oh wait, i needed to add "missing missing"? |
bingo
_________________
Nothing is secure / Security is always a trade-off with usability / Do not assume anything / Trust no-one, nothing / Paranoia is your friend / Think for yourself |
|
| Back to top |
|
|
bunder
Bodhisattva
Joined: 10 Apr 2004
Posts: 5947
|
| Posted: Thu May 07, 2009 10:30 am Post subject: |
|
|
finally got it going last night, but i still didn't have enough space.
woke up to this:
| Quote: | cp: writing `../md0/1056_20080321115900.nuv': No space left on device
|
luckily it's only a small handful of files, and i won't even mention that i accidentally blew away my movie partition. 
| Quote: | md1 : active raid5 hdd2[1] hdc2[0]
78236416 blocks level 5, 64k chunk, algorithm 2 [3/2] [UU_]
md0 : active raid5 hdd1[1] hdc1[0]
410155264 blocks level 5, 64k chunk, algorithm 2 [3/2] [UU_]
|
(how come they're backwards? eg: d then c)
at any rate, i'll have to find a spot for some of these files.
edit/update: i think we're good.
| Quote: | # cat /proc/mdstat
Personalities : [raid0] [raid1] [raid6] [raid5] [raid4]
md1 : active raid5 hdb2[3] hda2[2] hdd2[1] hdc2[0]
78236416 blocks super 0.91 level 5, 64k chunk, algorithm 2 [4/4] [UUUU]
resync=DELAYED
md0 : active raid5 hdb1[3] hda1[2] hdd1[1] hdc1[0]
410155264 blocks super 0.91 level 5, 64k chunk, algorithm 2 [4/4] [UUUU]
[>....................] reshape = 0.3% (635648/205077632) finish=665.7min speed=5116K/sec
unused devices: <none>
|
_________________
| Neddyseagoon wrote: | | The problem with leaving is that you can only do it once and it reduces your influence. |
banned from #gentoo since sept 2017 |
|
| Back to top |
|
|
|
|
You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum
|
|
Kernel & Hardware
