Hey List;
I have no experience with software RAIDs; at work we only use hardware RAIDs and I'm looking to implement, probably a RAID 5 set up at home for a media server however I have a few questions;
I have three 1TB drives in various places; one is inside a USB caddy, one is inside my PC and in is inside my existing media centre.
Is it possible to add these three drives to another one to give me 4TBs of space in a RAID setup without having to wipe the drives as in my experience which is only with hardware RAIDs, I have normally formatted all the disks before creating the RAID? The thing is, if that is the case I will need to transfer 3TB of stuff somewhere (I have an idea where, if this were the case), make the RAID then transfer it all back but I really don't want to do that as I'm sure you can imagine.
Also, if the above where possible; in the future could I then keep adding more drives and expanding the RAID?
Note: Obviously I know for this to be a RAID 5 I would need extra drives but the RAID level is undecided, but provisionally I think it will be RAID 5.
On a side note, I cobbled together my new media centre running Ubuntu but I might move it back to CentOS, it was originally CentOS and that is my favoured distro, but I would rather not now it is running Ubuntu happily I'm just wondering, is this all achievable in Ubuntu? Granted people on the CentOS mailing list might not know that, but if anyone knows that it is all achievable in CentOS then I would move back?
Thanks for reading.
Regards, James ;)
few things;
i'm not hip to the latest barney phife shareware utils that may allow some kind of dynamic raiding bs, but generally, no, you must format those drive to the desired fs after raiding them so back up all data u want housed (or hozed) on that new raid.
basic math is that for a raid 5, subtract 1 drives worth of space and thats your usable space.
so if u have 4x1tb drives, raw space is 3tb in a raid 5.
md (software raid mechanism or watever u wanna call it) allows u to grow so if your raid is formatted with an fs that allows u to grow, then u r good to go (lvm will allow u to grow).
On Sep 27, 2009, at 1:03 PM, James Bensley wrote:
Hey List;
I have no experience with software RAIDs; at work we only use hardware RAIDs and I'm looking to implement, probably a RAID 5 set up at home for a media server however I have a few questions;
I have three 1TB drives in various places; one is inside a USB caddy, one is inside my PC and in is inside my existing media centre.
Is it possible to add these three drives to another one to give me 4TBs of space in a RAID setup without having to wipe the drives as in my experience which is only with hardware RAIDs, I have normally formatted all the disks before creating the RAID? The thing is, if that is the case I will need to transfer 3TB of stuff somewhere (I have an idea where, if this were the case), make the RAID then transfer it all back but I really don't want to do that as I'm sure you can imagine.
Also, if the above where possible; in the future could I then keep adding more drives and expanding the RAID?
Note: Obviously I know for this to be a RAID 5 I would need extra drives but the RAID level is undecided, but provisionally I think it will be RAID 5.
On a side note, I cobbled together my new media centre running Ubuntu but I might move it back to CentOS, it was originally CentOS and that is my favoured distro, but I would rather not now it is running Ubuntu happily I'm just wondering, is this all achievable in Ubuntu? Granted people on the CentOS mailing list might not know that, but if anyone knows that it is all achievable in CentOS then I would move back?
Thanks for reading.
Regards, James ;)
--
Ted Turner - "Sports is like a war without the killing." - http://www.brainyquote.com/quotes/authors/t/ted_turner.html _______________________________________________ CentOS mailing list CentOS@centos.org http://lists.centos.org/mailman/listinfo/centos
2009/9/27 aurfalien@gmail.com:
you must format those drive to the desired fs after raiding them
Those where my initial thoughts, damn it! Hmm, going to have to bring a sand box server home from work, oh the hassle. Thanks for the confirmation though. ...
md (software raid mechanism or watever u wanna call it) allows u to grow so if your raid is formatted with an fs that allows u to grow, then u r good to go (lvm will allow u to grow).
This was also my current belief, again thanks for confirming that.
Regards James ;)
as a precaution, make a backup of your /etc/md.conf when all dun bulding the raid.
good luck.
i built a raid 6 using md as i had 40x1Tb drives to do.
i tried raid 5 with 3 or 4 hot spares but couldn't get it to work like i wanted.
i think its wise that for every 10 drives, do a hot spare for statistics sake.
On Sep 27, 2009, at 1:49 PM, James Bensley wrote:
2009/9/27 aurfalien@gmail.com:
you must format those drive to the desired fs after raiding them
Those where my initial thoughts, damn it! Hmm, going to have to bring a sand box server home from work, oh the hassle. Thanks for the confirmation though. ...
md (software raid mechanism or watever u wanna call it) allows u to grow so if your raid is formatted with an fs that allows u to grow, then u r good to go (lvm will allow u to grow).
This was also my current belief, again thanks for confirming that.
Regards James ;) --
Stephen Leacock - "I detest life-insurance agents: they always argue that I shall some day die, which is not so." - http://www.brainyquote.com/quotes/authors/s/stephen_leacock.html _______________________________________________ CentOS mailing list CentOS@centos.org http://lists.centos.org/mailman/listinfo/centos
On Sun, 2009-09-27 at 21:49 +0100, James Bensley wrote:
2009/9/27 aurfalien@gmail.com:
you must format those drive to the desired fs after raiding them
Those where my initial thoughts, damn it! Hmm, going to have to bring a sand box server home from work, oh the hassle. Thanks for the confirmation though. ...
md (software raid mechanism or watever u wanna call it) allows u to grow so if your raid is formatted with an fs that allows u to grow, then u r good to go (lvm will allow u to grow).
This was also my current belief, again thanks for confirming that.
---- you will not be able to add more physical drives to an existing RAID-5 array. You can however, add new available storage space (RAID or drives) to an existing LVM partition with some of the LVM commands.
Just so we are clear...
RAID is physical drives. LVM is a partitioning scheme.
Craig
craig, u can grow via md which is what u do first, then grow the fs.
ex;
mdadm --grow ... to add a drive
then,
mdadm --add ... to gorw the raid onto it
then,
pvresize ... to grow your filesystem onto it.
atleast this is how i've dun it.
On Sep 27, 2009, at 2:07 PM, Craig White wrote:
On Sun, 2009-09-27 at 21:49 +0100, James Bensley wrote:
2009/9/27 aurfalien@gmail.com:
you must format those drive to the desired fs after raiding them
Those where my initial thoughts, damn it! Hmm, going to have to bring a sand box server home from work, oh the hassle. Thanks for the confirmation though. ...
md (software raid mechanism or watever u wanna call it) allows u to grow so if your raid is formatted with an fs that allows u to grow, then u r good to go (lvm will allow u to grow).
This was also my current belief, again thanks for confirming that.
you will not be able to add more physical drives to an existing RAID-5 array. You can however, add new available storage space (RAID or drives) to an existing LVM partition with some of the LVM commands.
Just so we are clear...
RAID is physical drives. LVM is a partitioning scheme.
Craig
-- This message has been scanned for viruses and dangerous content by MailScanner, and is believed to be clean.
CentOS mailing list CentOS@centos.org http://lists.centos.org/mailman/listinfo/centos
aurfalien@gmail.com wrote:
craig, u can grow via md which is what u do first, then grow the fs.
ex;
mdadm --grow ... to add a drive
He meant --add or -a
then,
mdadm --add ... to gorw the raid onto it
He meant --grow + changes in size/numofdevices/yadayada
then,
pvresize ... to grow your filesystem onto it.
atleast this is how i've dun it.
md (software raid mechanism or watever u wanna call it) allows u to grow so if your raid is formatted with an fs that allows u to grow, then u r good to go (lvm will allow u to grow).
This was also my current belief, again thanks for confirming that.
you will not be able to add more physical drives to an existing RAID-5
You can since 2007 (mainline kernel 2.6.17) and therefore should be available in Centos 5.
On 27/09/09 21:32, aurfalien@gmail.com wrote:
few things;
i'm not hip to the latest barney phife shareware utils that may allow some kind of dynamic raiding bs, but generally, no, you must format those drive to the desired fs after raiding them so back up all data u want housed (or hozed) on that new raid.
This is not true. Even with the extremely basic process, you only need to have 1 disk in 'being prepd' state - which means you can, if there is time and resources, shovel data around on the disks as they get used into the raid setup. Ofcourse, you will need to have 1 disk clear.
Eg. to convert this setup into a raid-5 : start with a 1 disk raid-1 degraded, move a disks worth of data onto there, bring in the second disk and let them sync into a 'normal' raid1 state, then convert that into a raid-5 with the 3rd disk. then bring in the 4th disk as hotspare.
with TiB's worth of data, its not going to be quick.
- KB
On Sep 27, 2009, at 8:56 PM, Karanbir Singh mail-lists@karan.org wrote:
On 27/09/09 21:32, aurfalien@gmail.com wrote:
few things;
i'm not hip to the latest barney phife shareware utils that may allow some kind of dynamic raiding bs, but generally, no, you must format those drive to the desired fs after raiding them so back up all data u want housed (or hozed) on that new raid.
This is not true. Even with the extremely basic process, you only need to have 1 disk in 'being prepd' state - which means you can, if there is time and resources, shovel data around on the disks as they get used into the raid setup. Ofcourse, you will need to have 1 disk clear.
Eg. to convert this setup into a raid-5 : start with a 1 disk raid-1 degraded, move a disks worth of data onto there, bring in the second disk and let them sync into a 'normal' raid1 state, then convert that into a raid-5 with the 3rd disk. then bring in the 4th disk as hotspare.
with TiB's worth of data, its not going to be quick.
Can you convert a RAID0 to a RAID5?
With a proper backup of the data in place that might be more efficient then a RAID1 to 5.
-Ross
Ross Walker wrote:
On Sep 27, 2009, at 8:56 PM, Karanbir Singh mail-lists@karan.org wrote:
On 27/09/09 21:32, aurfalien@gmail.com wrote:
few things;
i'm not hip to the latest barney phife shareware utils that may allow some kind of dynamic raiding bs, but generally, no, you must format those drive to the desired fs after raiding them so back up all data u want housed (or hozed) on that new raid.
This is not true. Even with the extremely basic process, you only need to have 1 disk in 'being prepd' state - which means you can, if there is time and resources, shovel data around on the disks as they get used into the raid setup. Ofcourse, you will need to have 1 disk clear.
Eg. to convert this setup into a raid-5 : start with a 1 disk raid-1 degraded, move a disks worth of data onto there, bring in the second disk and let them sync into a 'normal' raid1 state, then convert that into a raid-5 with the 3rd disk. then bring in the 4th disk as hotspare.
with TiB's worth of data, its not going to be quick.
Can you convert a RAID0 to a RAID5?
According to Neil Brown, yes.
On 28/09/09 02:12, Christopher Chan wrote:
Can you convert a RAID0 to a RAID5?
According to Neil Brown, yes.
erm, that would be quite interesting, since you cant split the raid-0 without first moving data into some state of sanity first. Can you point out where Neil says its something that mdraid can do at the moment ?
- KB
Karanbir Singh wrote:
On 28/09/09 02:12, Christopher Chan wrote:
Can you convert a RAID0 to a RAID5?
According to Neil Brown, yes.
erm, that would be quite interesting, since you cant split the raid-0 without first moving data into some state of sanity first. Can you point out where Neil says its something that mdraid can do at the moment ?
Ah sorry, did not read his blog carefully enough.
http://neil.brown.name/blog/20090817000931
He had a line about RAID0 to RAID5 when I skimmed over it. Looking again...it is not implemented yet.
so you're giving advice based on what you're read?
know that my advice is stuff i've actually done.
On Sep 27, 2009, at 6:26 PM, Christopher Chan wrote:
Karanbir Singh wrote:
On 28/09/09 02:12, Christopher Chan wrote:
Can you convert a RAID0 to a RAID5?
According to Neil Brown, yes.
erm, that would be quite interesting, since you cant split the raid-0 without first moving data into some state of sanity first. Can you point out where Neil says its something that mdraid can do at the moment ?
Ah sorry, did not read his blog carefully enough.
http://neil.brown.name/blog/20090817000931
He had a line about RAID0 to RAID5 when I skimmed over it. Looking again...it is not implemented yet. _______________________________________________ CentOS mailing list CentOS@centos.org http://lists.centos.org/mailman/listinfo/centos
aurfalien@gmail.com wrote:
so you're giving advice based on what you're read?
1) You do not use --grow to add a device.
2) You do not use --add to grow a device
3) I did not give advice
know that my advice is stuff i've actually done.
Oh no doubt. Does that mean you do not make mistakes?
BTW, the question was: Can you convert a RAID0 to a RAID5? and the answer is YES from both Neil Brown and Karanbir's post. The only part where I was wrong was I thought it was already implemented. If you want to call that 'giving advice' then hey, feel free.
On Sep 27, 2009, at 6:26 PM, Christopher Chan wrote:
Karanbir Singh wrote:
On 28/09/09 02:12, Christopher Chan wrote:
Can you convert a RAID0 to a RAID5?
According to Neil Brown, yes.
erm, that would be quite interesting, since you cant split the raid-0 without first moving data into some state of sanity first. Can you point out where Neil says its something that mdraid can do at the moment ?
Ah sorry, did not read his blog carefully enough.
http://neil.brown.name/blog/20090817000931
He had a line about RAID0 to RAID5 when I skimmed over it. Looking again...it is not implemented yet. _______________________________________________ CentOS mailing list CentOS@centos.org http://lists.centos.org/mailman/listinfo/centos
CentOS mailing list CentOS@centos.org http://lists.centos.org/mailman/listinfo/centos
yea, it means u never done it yourself.
On Sep 27, 2009, at 7:02 PM, Christopher Chan wrote:
aurfalien@gmail.com wrote:
so you're giving advice based on what you're read?
You do not use --grow to add a device.
You do not use --add to grow a device
I did not give advice
know that my advice is stuff i've actually done.
Oh no doubt. Does that mean you do not make mistakes?
BTW, the question was: Can you convert a RAID0 to a RAID5? and the answer is YES from both Neil Brown and Karanbir's post. The only part where I was wrong was I thought it was already implemented. If you want to call that 'giving advice' then hey, feel free.
On Sep 27, 2009, at 6:26 PM, Christopher Chan wrote:
Karanbir Singh wrote:
On 28/09/09 02:12, Christopher Chan wrote:
Can you convert a RAID0 to a RAID5?
According to Neil Brown, yes.
erm, that would be quite interesting, since you cant split the raid-0 without first moving data into some state of sanity first. Can you point out where Neil says its something that mdraid can do at the moment ?
Ah sorry, did not read his blog carefully enough.
http://neil.brown.name/blog/20090817000931
He had a line about RAID0 to RAID5 when I skimmed over it. Looking again...it is not implemented yet. _______________________________________________ CentOS mailing list CentOS@centos.org http://lists.centos.org/mailman/listinfo/centos
CentOS mailing list CentOS@centos.org http://lists.centos.org/mailman/listinfo/centos
CentOS mailing list CentOS@centos.org http://lists.centos.org/mailman/listinfo/centos
aurfalien@gmail.com wrote:
yea, it means u never done it yourself.
Yes, I do not do raid5. That does not mean I have not used mdadm. I do raid1+0 (not raid10) and I do not see how that disqualifies me from pointing out that you have the wrong flags for the intended result.
I also do raidz and will probably grow those arrays when support for that is added to OpenSolaris.
Likewise, I have used 3ware and know the difference between using 3ware 750x/850x for raid5 and using 3ware boards that come with bbu caches.
I have done software raid and hardware raid for clusters of servers for a sizable messaging service provider now owned by IBM.
I do not know what makes you think that you can be so full of yourself to the point of thinking that you do not ever make mistakes.
On Sep 27, 2009, at 7:02 PM, Christopher Chan wrote:
aurfalien@gmail.com wrote:
so you're giving advice based on what you're read?
You do not use --grow to add a device.
You do not use --add to grow a device
I did not give advice
know that my advice is stuff i've actually done.
Oh no doubt. Does that mean you do not make mistakes?
BTW, the question was: Can you convert a RAID0 to a RAID5? and the answer is YES from both Neil Brown and Karanbir's post. The only part where I was wrong was I thought it was already implemented. If you want to call that 'giving advice' then hey, feel free.
On Sep 27, 2009, at 6:26 PM, Christopher Chan wrote:
Karanbir Singh wrote:
On 28/09/09 02:12, Christopher Chan wrote:
> Can you convert a RAID0 to a RAID5? > > > According to Neil Brown, yes.
erm, that would be quite interesting, since you cant split the raid-0 without first moving data into some state of sanity first. Can you point out where Neil says its something that mdraid can do at the moment ?
Ah sorry, did not read his blog carefully enough.
http://neil.brown.name/blog/20090817000931
He had a line about RAID0 to RAID5 when I skimmed over it. Looking again...it is not implemented yet. _______________________________________________ CentOS mailing list CentOS@centos.org http://lists.centos.org/mailman/listinfo/centos
CentOS mailing list CentOS@centos.org http://lists.centos.org/mailman/listinfo/centos
CentOS mailing list CentOS@centos.org http://lists.centos.org/mailman/listinfo/centos
CentOS mailing list CentOS@centos.org http://lists.centos.org/mailman/listinfo/centos
Hi All,
Thanks for all your replies. I understand what I have to do now and have read several tutorials to get a good grasp of everything.
All your input has been greatly appreciated. Thank you all.
Regards, James ;)
On 28/09/09 02:11, Ross Walker wrote:
Eg. to convert this setup into a raid-5 : start with a 1 disk raid-1 degraded, move a disks worth of data onto there, bring in the second disk and let them sync into a 'normal' raid1 state, then convert that into a raid-5 with the 3rd disk. then bring in the 4th disk as hotspare.
with TiB's worth of data, its not going to be quick.
Can you convert a RAID0 to a RAID5?
How would you create a 1 disk degraded raid0 :) well, you could - but you wont get any data onto it! So you would need to start with 2 disks, and with that you then need to store 2 disks worth of stuff somewhere while this is building[1].
- KB
[1]: there are h/w based cards that are known to convert a raid-0 to a raid-1 and then onto a raid-5 and then 6. so there are ways and means to implement it.
On Sep 27, 2009, at 9:35 PM, Karanbir Singh mail-lists@karan.org wrote:
On 28/09/09 02:11, Ross Walker wrote:
Eg. to convert this setup into a raid-5 : start with a 1 disk raid-1 degraded, move a disks worth of data onto there, bring in the second disk and let them sync into a 'normal' raid1 state, then convert that into a raid-5 with the 3rd disk. then bring in the 4th disk as hotspare.
with TiB's worth of data, its not going to be quick.
Can you convert a RAID0 to a RAID5?
How would you create a 1 disk degraded raid0 :) well, you could - but you wont get any data onto it! So you would need to start with 2 disks, and with that you then need to store 2 disks worth of stuff somewhere while this is building[1].
Can't you start with a single disk raid0 and expand? I think that is possible no? Then with a two disk raid0 add a parity disk (raid4) then distribute the parity for a raid5.
Of course if Linux RAID doesn't support that then it can't be done yet, but it would be possible.
-Ross
Can you say;
Use a sledge hammer to pound in a nail?
While its nice to expand ones technique, regardless of time and resources, the shortest distance between 2 points is a straight line.
On Sep 27, 2009, at 5:56 PM, Karanbir Singh wrote:
On 27/09/09 21:32, aurfalien@gmail.com wrote:
few things;
i'm not hip to the latest barney phife shareware utils that may allow some kind of dynamic raiding bs, but generally, no, you must format those drive to the desired fs after raiding them so back up all data u want housed (or hozed) on that new raid.
This is not true. Even with the extremely basic process, you only need to have 1 disk in 'being prepd' state - which means you can, if there is time and resources, shovel data around on the disks as they get used into the raid setup. Ofcourse, you will need to have 1 disk clear.
Eg. to convert this setup into a raid-5 : start with a 1 disk raid-1 degraded, move a disks worth of data onto there, bring in the second disk and let them sync into a 'normal' raid1 state, then convert that into a raid-5 with the 3rd disk. then bring in the 4th disk as hotspare.
with TiB's worth of data, its not going to be quick.
- KB
CentOS mailing list CentOS@centos.org http://lists.centos.org/mailman/listinfo/centos
On 28/09/09 02:19, aurfalien@gmail.com wrote:
Can you say;
Use a sledge hammer to pound in a nail?
While its nice to expand ones technique, regardless of time and resources, the shortest distance between 2 points is a straight line.
Thats a commpletely irrelevant statement. No matter what instrument you use to drive a nail, your line of impact and nail penetration is going to be the same. So not only is your analagy irrelevant its also not really an analagy.
Anyway, your original statement seemed to imply that it wasent possible or even recommended to move storage shape without wiping all data. Which is not correct, which ever way you want to play it.
Also, your email client is broken. You seem unable to either quote properly or bottom post ( both of which are requested from posters to this list )
- KB
have you tried your advice cause it sounds like a bunch of theory.
give advice you've actually tried instead of sounding like some academic.
back it up and build the raid from scratch instead of some theory non sense.
On Sep 27, 2009, at 6:26 PM, Karanbir Singh wrote:
On 28/09/09 02:19, aurfalien@gmail.com wrote:
Can you say;
Use a sledge hammer to pound in a nail?
While its nice to expand ones technique, regardless of time and resources, the shortest distance between 2 points is a straight line.
Thats a commpletely irrelevant statement. No matter what instrument you use to drive a nail, your line of impact and nail penetration is going to be the same. So not only is your analagy irrelevant its also not really an analagy.
Anyway, your original statement seemed to imply that it wasent possible or even recommended to move storage shape without wiping all data. Which is not correct, which ever way you want to play it.
Also, your email client is broken. You seem unable to either quote properly or bottom post ( both of which are requested from posters to this list )
- KB
CentOS mailing list CentOS@centos.org http://lists.centos.org/mailman/listinfo/centos
On 28/09/09 02:52, aurfalien@gmail.com wrote:
have you tried your advice cause it sounds like a bunch of theory.
give advice you've actually tried instead of sounding like some academic.
because you are unaware of things does not mean it does not happen. I've reshaped raid's dozens of time and do it as a matter of regular practise as and when required. I suggest you spend a bit of time making sure you know what you are talking about before actually making authoritative statements.
back it up and build the raid from scratch instead of some theory non sense.
Were meeting on the 29th in london[1] for drinks, feel free to come over and I'll be happy to demonstrate this whole process to you.
- KB
[1]: 6:30 pm onwards at the King and Queens, 49 Cleveland Street - its by the shadow of the BT tower. You wont find it hard to spot the CentOS crowd.
You have me there.
I've only done perhaps 6 sw based Raids because I would only use hw Raids for production data, so I sleep better at night.
Tell me, you ever support your Frankenstein contraptions?
A kind invite and I'm sure we would have a good time talking shop but I'm not in Europe these days.
On Sep 27, 2009, at 7:07 PM, Karanbir Singh wrote:
On 28/09/09 02:52, aurfalien@gmail.com wrote:
have you tried your advice cause it sounds like a bunch of theory.
give advice you've actually tried instead of sounding like some academic.
because you are unaware of things does not mean it does not happen. I've reshaped raid's dozens of time and do it as a matter of regular practise as and when required. I suggest you spend a bit of time making sure you know what you are talking about before actually making authoritative statements.
back it up and build the raid from scratch instead of some theory non sense.
Were meeting on the 29th in london[1] for drinks, feel free to come over and I'll be happy to demonstrate this whole process to you.
- KB
[1]: 6:30 pm onwards at the King and Queens, 49 Cleveland Street - its by the shadow of the BT tower. You wont find it hard to spot the CentOS crowd. _______________________________________________ CentOS mailing list CentOS@centos.org http://lists.centos.org/mailman/listinfo/centos
On 28/09/09 04:20, aurfalien@gmail.com wrote:
You have me there.
I've only done perhaps 6 sw based Raids because I would only use hw Raids for production data, so I sleep better at night.
Tell me, you ever support your Frankenstein contraptions?
I am not sure what h/w raid you are using - almost every major hba developer has had storage shaping since the last 4 - 7 years. Specially on self contained storage setups.
And yes, I totally support mdraid/areca/3ware/lsi/megaraid/servraid/cciss sort of solutions - each one of them has raid migrations.
my feeling is that you need to spend a bit of time and see how the storage world is moving these days :)
- KB
To me, a storage developer and there technique is very different then using plain old mdadm for sw raids within a Linux distro.
I know vendors like Bluearc, Netapp, etc spin there own firmware, ie software but thats very very diff then mdadm within the OS.
And yes, i could do for more reading for sure.
On Sep 27, 2009, at 8:28 PM, Karanbir Singh wrote:
On 28/09/09 04:20, aurfalien@gmail.com wrote:
You have me there.
I've only done perhaps 6 sw based Raids because I would only use hw Raids for production data, so I sleep better at night.
Tell me, you ever support your Frankenstein contraptions?
I am not sure what h/w raid you are using - almost every major hba developer has had storage shaping since the last 4 - 7 years. Specially on self contained storage setups.
And yes, I totally support mdraid/areca/3ware/lsi/megaraid/servraid/cciss sort of solutions - each one of them has raid migrations.
my feeling is that you need to spend a bit of time and see how the storage world is moving these days :)
- KB
CentOS mailing list CentOS@centos.org http://lists.centos.org/mailman/listinfo/centos
On 28/09/2009, at 5:33 AM, James Bensley wrote:
Hey List;
I have no experience with software RAIDs; at work we only use hardware RAIDs and I'm looking to implement, probably a RAID 5 set up at home for a media server however I have a few questions;
I have three 1TB drives in various places; one is inside a USB caddy, one is inside my PC and in is inside my existing media centre.
Is it possible to add these three drives to another one to give me 4TBs of space in a RAID setup without having to wipe the drives as in my experience which is only with hardware RAIDs, I have normally formatted all the disks before creating the RAID? The thing is, if that is the case I will need to transfer 3TB of stuff somewhere (I have an idea where, if this were the case), make the RAID then transfer it all back but I really don't want to do that as I'm sure you can imagine.
Also, if the above where possible; in the future could I then keep adding more drives and expanding the RAID?
Note: Obviously I know for this to be a RAID 5 I would need extra drives but the RAID level is undecided, but provisionally I think it will be RAID 5.
On a side note, I cobbled together my new media centre running Ubuntu but I might move it back to CentOS, it was originally CentOS and that is my favoured distro, but I would rather not now it is running Ubuntu happily I'm just wondering, is this all achievable in Ubuntu? Granted people on the CentOS mailing list might not know that, but if anyone knows that it is all achievable in CentOS then I would move back?
Thanks for reading.
Regards, James ;)
Hi James,
I'm going to try and answer all your questions concisely because the main thread seems to have turned into an debate more than anything else.
Someone has already stated that the total capacity of a RAID5 array is that of all disks minus 1. So with your four 1TB drives you can have 3TB of space.
Moving all the data elsewhere is probably the easiest option to take and also the fastest. You could in theory create a degraded RAID5 array with two drives then migrate data across while growing the array at the same time then growing the array each time all data from a drive has been moved. This would take very long, require a lot of work, and I would therefore recommend against it.
Down the track you can add more drives to a RAID5 array and "grow" the array. This process also takes a long time. You do not need to be using LVM on top of mdadm to complete this. Once an array has finished growing you can use resize2fs to increase the size of an ext3 file system. It can even be done with the filesystem mounted, though personally I think it's best to do a forced fsck.ext3 on the file system then run the resize2fs on it while it is offline.
The process to grow a RAID5 array is to add a new drive to the array (which adds it as a hot spare) then run a --grow including the --raid- devices=5 argument (which is what you'd use if you were to expand your 4 drive RAID5 system to a 5 drive array down the track).
If you were using LVM on top you'd need to use pvresize and then lvresize to increase the size of your logical volumes, and then resize the file system.
One thing I would strong suggest is that you get a UPS for your server. In my experience RAID5 is much more prone to problems if you experience power loss, compared to RAID1 or even a single disk system.
I hope that helps.
Regards, Oliver
--
Ted Turner - "Sports is like a war without the killing." - http://www.brainyquote.com/quotes/authors/t/ted_turner.html _______________________________________________ CentOS mailing list CentOS@centos.org http://lists.centos.org/mailman/listinfo/centos
-
aurfalien＠gmail.com -
Christopher Chan -
Craig White -
James Bensley -
Karanbir Singh -
Oliver Ransom -
Ross Walker