Replacing (4) HDDs in RAID 10 with (4) SSDs

SergioM · Jan 6, 2016

I have a server that has (4) 500GB HDDs. I want to replace the drives with (4) 960GB SSDs. Ideally I would like to do it without installing a new OS on the new drives. I think one of two ways should work.

(1) Replace 1 drive at a time. Each time a new drive is introduced the data will propagate until all 4 drives are done.

(2) Shut down the computer & clone all 4 drives. At once connect all 4 drives and boot the computer.

I'm not actually sure about either idea, but both seem plausible & this is pretty important. Does anyone have experience with this?

kanewolf · Jan 6, 2016

Adding a 960GB SSD into a RAID with 3 500GB disks will only use 500GB of the 960GB. You will do better to backup the existing RAID, build a new RAID set with the SSDs using the full size and restore the backup.

13thmonkey · Jan 6, 2016

do you really need to raid the SSD's?

problematiq · Jan 6, 2016

If you can locate a spare 1TB HDD then use something like Acronis to do a copy to the 1TB drive then remove the array and build a new array using the SSD's, then copy the data back to the raid. Depending on the value of the information you might switch to a raid 5.

How many drive bay's do you have?

problematiq · Jan 6, 2016

13thmonkey :

He may want a solid drive without multiple partitions.

13thmonkey · Jan 6, 2016

in which case take 1 of SSD's dump all of the data on that (4 500's in R10 should be 1TB), then make an array with the remaining 3 as jbod, move it and add the 4th drive.

SergioM · Jan 6, 2016

13thmonkey :

Yes

13thmonkey · Jan 6, 2016

SergioM :

for what benefit?

SergioM · Jan 6, 2016

13thmonkey :

It's a mission critical server which hosts a custom software & database that the entire company uses. The DB & respective files are backed up, but downtime of a day or two would be pretty costly.

13thmonkey · Jan 8, 2016

SergioM :

And you are asking here?
and you can't afford temporary storage?
or a SAN?
or a backup mechanism?
you do understand that YOU are responsible if this does not work, not us.

SergioM · Jan 11, 2016

I came here for technical help, not a lecture on responsibility.

If you are unwilling to help with me learn about this, then don't. But keep your soapbox tucked away.

kanewolf · Jan 11, 2016

SergioM :

Technical help is PROCESS as well as HW. Best practices for a business critical DB would be to have a second server which could either fail-over or be used as a cluster. RAID helps in some failure scenarios, but it doesn't sound like best practices are being implemented. Journal files from the DB on different physical volumes from the DB files, OS separate from everything ...

There are plenty of lessons to be learned besides how to grow a RAID volume.

SergioM · Jan 11, 2016

kanewolf :

I hear that and agree.

I didn't quote the reply but my gripe was with 13thmonkey specifically. It was unusual for him to need to know exactly what I was doing with the server. Then when I volunteered that information, I didn't appreciate the needless lecturing.

As for the process, I recognize that a second & equally powerful server would be optimal. Though I'm sure that I'm not the only one in the forum who exists in a job where resources are limited & my boss does not understand and will not supply the server regardless of the consequences.

To that end, the advice that I need is on how to replace 4 HDDs from a RAID configuration to SSD as safely as possible. This may include backups and other precautionary steps. However I am not confused about backing up software or other precautionary steps. I am uncertain as to what the best practice is fro replacing HDDs in a RAID configuration & that is why I came here for help.

kanewolf · Jan 11, 2016

SergioM :

kanewolf :

I hear that and agree.

I didn't quote the reply but my gripe was with 13thmonkey specifically. It was unusual for him to need to know exactly what I was doing with the server. Then when I volunteered that information, I didn't appreciate the needless lecturing.

As for the process, I recognize that a second & equally powerful server would be optimal. Though I'm sure that I'm not the only one in the forum who exists in a job where resources are limited & my boss does not understand and will not supply the server regardless of the consequences.

To that end, the advice that I need is on how to replace 4 HDDs from a RAID configuration to SSD as safely as possible. This may include backups and other precautionary steps. However I am not confused about backing up software or other precautionary steps. I am uncertain as to what the best practice is fro replacing HDDs in a RAID configuration & that is why I came here for help.

Which is EXACTLY where I started ... Back it up create a brand new RAID set and reload the backup ....

13thmonkey · Jan 12, 2016

SergioM :

OK, the reason I asked was that we get large numbers of people on here who want to raid SSD's because they think it'll be quicker for normal home use. So it's a perfectly reasonable question to help people not waste money on an un-needed solution.

To do it as safely as possible, back up and restore. Simple. To bodge it, well that's your choice.

Of my 4 questions, 3 of them point to solutions, 1 of them asks if you are professional (which you seem to be saying that you are) then why are you asking on an enthusiast website? again a perfectly reasonable question, you could be volunteering to look after a charities systems etc. But you seem to be in a for-profit organisation. My fifth statement is reminding you that you are responsible, there is no comeback on us, what you are doing and the way you want to go about it is risky.

For the size of solution you have just buy a 1TB USB drive as a backup, take the drives put them safe somewhere until you have it rebuilt. If your boss refuses the £50 for that I'd buy it yourself for your own peace of mind. Yes there will be downtime, work sunday or some other down time.

SergioM · Jan 12, 2016

kanewolf :

I accidentally marked another answer as best response, but this is probably my best bet. The upside is that it allows me to simultaneously connect 4 new SSDs without risking the integrity of the current set.

The downside is that without cloning (or an equivalent) I have to pay the company to reinstall the software/services. I would really prefer to replace the SSDs & have the computer reinitialize (or to clone) but if this is the only safe way, then there it is. Thanks

13thmonkey :

Firstly, you're wrong. The software is custom, doesn't come with an installer & is dependent on windows services & file locations & names that aren't just plug & play. If I don't clone the drives (or it's equivalent) then I have to pay this company to reinstall & transfer from my backup location.

Secondly, I understand that you can't hear (or won't acknowledge) your own attitude. Frankly I feel as uncomfortable speaking to you about basic human interactions as I do having to hear your comments. Put simply, thank you for your help, but I prefer to do without it.

13thmonkey · Jan 12, 2016

I've de-selected the best answer.

And by backing up and restoring the whole drive all settings are maintained, that's what a backup does. I never mentioned installing or reinstalling the software.

As to your 'discomfort', I'm not really sure what I've said to create that, other than the truth.

jimmysmitty · Jan 12, 2016

I was not planning on commenting on this as the solution has already been presented but I do have a few questions that did pop up.

1. If this is a critical database and software, why are you running a RAID10 instead of RAID5 or better yet using 5 drives and RAID6 for two parity drives?

2. For the same as above why are you using SSDs for a database that probably has hundreds if not more reads/writes per day when SSDs current limitation is in P/E cycles?

3. What SSD model are these specifically? You are not using consumer grade SSDs in an enterprise class environment are you?

Now you can take offense to the questions, which I hope not, but I am trying to get as much information as possible. I can answer all of those questions simply:

1. You should be using RAID5 or 6 for critical databases for the best redundancy. I work for a nation wide aggregate company (one of the top 25 in the US) and we use RAID5 in all of our systems (moving to RAID6 now) with multiple backups and redundancy of all of our setups and we had a catastrophic failure on one where, and this is rare, we had three drives fail.

2/3. While SSDs are nice for their performance they are limited by their P/E cycles. Modern ones do have higher P/E cycles but not good enough for server class use. That is why they have enterprise class SSDs with much higher P/E cycles than consumer.

Again this is just to help you out, as were the questions from 13thMonkey, as the more information we have the better we can help.

SergioM · Jan 14, 2016

jimmysmitty :

1) I was told that RAID5 would give me better read performance but not better write performance. In my cases I needed both. Also, I assumed that RAID10 would give me sufficient safety. RAID6 is out of the question because my server Dell T110 II has 4 drive bays. I also was told (though I cannot find info to confirm today) that the PERC S110 would only support 4 drives.

Do you still suggest RAID5 over RAID10 knowing that lots of writing will occur & that the performance of the software is pretty slow?

2) The software is custom but unbelievably slow. SSD would offer an increase in performance.

3) Here I may be making a mistake. I have (2) 1TB enterprise class SSDs waiting (bought by somone else) but my boss has insisted that he doesn’t want to spend on another 2. As an olive branch, I offered 4x 500gb consumer grade SSDs assuming that I could just swap them out upon failure since it’s a RAID configuration. Though I don’t actually know how long it will take before one fails under the load. Swapping drives in the past has been quick and easy, but I also don’t know if it’s always that quick or easy

Would you suggest that I stick with HDDs in RAID, 4 consumer SSDs or switch over to enterpreise SSDs with just 2 drives?

I can only imagine how I came off if it seems that I am offended by literally anything, but I am grateful that you’ve offered the advice.

jimmysmitty · Jan 14, 2016

You are correct on your PERC 110. It only supports up to 4 HDDs which limits the RAID options to 0, 1, 5 or 10.

I understand the software is slow but redundancy is of the utmost importance when it is company critical, so sacrificing some write performance is not a bad idea.

Since you have two enterprise class SSDs right now it would probably be best to get them into a RAID 1 for redundancy as the read/write performance of a SSD is so much greater than a HDD (especially in IOPS) that even a 20% performance loss would still blow your RAID 10 away.

Preferably you should get the owner to buy two more to do a RAID 5 because a RAID 5 offers better performance than a RAID 1 but still has parity. I would not do consumer grade SSDs. The problem is that if you compare them, there is a big difference between consumer and enterprise SSDs:

http://ark.intel.com/compare/86741,82937

The 750 is the consumer drive, the 3710 is the enterprise drive. Both are rated for the same days of writes but the enterprise drive is capable of 24.3 PetaBytes of writes per day while the consumer drive is capable of up to 70GB of writes per day.

The easy way to put it to the man that makes the decision is this: We can do consumer drives in a RAID 5 that will have a much higher failure rate and thus cost more money. Or we do 4 enterprise drives in RAID 5 with much better reliability costing less in the long run (less need to replace drives).

Replacing (4) HDDs in RAID 10 with (4) SSDs

Distinguished

Titan

Titan

Reputable

Reputable

Titan

Distinguished

Titan

Distinguished

Titan

Distinguished

Titan

Distinguished

Titan

Titan

Distinguished

Titan

Champion

Distinguished

Champion

Share this page