Replacing smaller disks with larger disks

[thestraycat]

I'm pretty sure i know the answer to this question but thought i'd get confirmation regardless.

 

I have an array of:

 

20 x 2TB disks

2 x 6TB parity disks.

 

I plan on swapping out 13 of the 2tb disks with 6tb disks.

 

Can i pull 2 disks at a time as i'm running 2 disks of parity?

Would the risk not be the same as someone running 1 disk of parity swapping out 1 disk at a time?

 

I have an inkling that the answer may be: "if there were an issue during the rebuild you'd stand to potentially lose 2 disks of data rather than 1." or similar.

Edited March 4, 2022 by thestraycat
adding

[JonathanM]

Assuming all the disks are healthy, the risks of rebuilding 2 at a time are pretty low, especially considering you still have the disks you are replacing, and the data on those should be fine should something go sideways.

[thestraycat]

@JonathanM - This is what i thought too. I take it i'd be vulnerable of losing any additional disk that failed during the 2 disk rebuild? (and having to put back the old 2tb disk to replace the data) In theory it should take the same time to rebuild 2 disks as it would 1 disk right?

 

Can anyone that's done it vouch for the 2 disk approach working?

Edited March 4, 2022 by thestraycat

[Frank1940]

Do you really need an additional 52TB of storage at this exact point in time.  The reason that I mention it is that a purchase of that many HD's at one time from a single source would mean that all of the disks would be coming from a single manufacturing lot.  (If that lot would have a manufacturing defect in it, you could be looking at a lot of problems in the future.  You could make an argument that you could be buying into a lot of superior quality.  However, by spreading out the purchases over a longer period of time, you would be averaging out to the quality level of the design itself.  If the design itself were found to have a serious design/manufacturing problem, you might be alerted to it and make a switch to another manufacturer and/or model.)

[JonathanM]

Frank has a really good point. I would not add all the disks at once, rather a staggered strategy, where you replace the worst of the old disks as you actually need the free space.

 

Every time either

a. any disk shows signs of weakness,

or

b. total array free space dips below 6TB,

 

I'd replace a disk until both a. and b. are no longer an issue.

 

That way, at least the new disks would have a varied amount of time and use, and if you got a dud among the 6TB's, you could immediately replace it with the next 6TB in your stack of spares.

 

It's always a good idea to keep 1 good drive the same size as parity or larger in your back pocket as a ready spare, and replacing known good drives with unknown new drives when you don't need the space is adding risk you don't need.

 

Another point to consider is possibly lowering your spindle count, but that is risky with Unraid, it's not really designed with drive removal in mind so you have to do some gyrations to accomplish it without losing parity protection for the duration of the removal.

[thestraycat]

@Frank1940 @JonathanM

 

These are all good point, but i feel they've all been factored in to my plan already so didn't go into detail on this post as i wanted to keep it focused on just the risks and capability of unraid to handle replacing multiple disks at once. Looking for clarification that i can in fact replace 2 disks at once and what the risks of this are. 

 

"Do you really need an additional 52TB of storage at this exact point in time."

>> Basically yes. My array is 91% full and my long winded plan (that i didnt post as to save a 'block of text' question that got no replies) was to always hold a disk or two back for emergency replacement.

 

"The reason that I mention it is that a purchase of that many HD's at one time from a single source would mean that all of the disks would be coming from a single manufacturing lot."

>> The disks are used, SMART tested as 100% healthy, with staggered data written and time powered on between them all. And from different batches and have different manafactured dates. That's about all the risk i could mitigate when purchasing them.

 

"You could make an argument that you could be buying into a lot of superior quality. "

>> I feel i am. As these are battle tested with lower usage and with potential warranties to utilize. 

 

"Frank has a really good point. I would not add all the disks at once, rather a staggered strategy, where you replace the worst of the old disks as you actually need the free space."

>> Yup this has always been my plan. Replace both parity disks for 6tb disks, then get my current array usage of about 36tb onto 6tb disks add 2 more 6tb disks to increase headroom of the array and then save 1 to 2 back.. Mainly for electricity savings, heat savings (as i have a dense 24 bay 4u rack),  putting on needless mileage on unused harddisks And also to have the security of future replacements disks on hand. Lastly, the 2TB disks all need removing as they now need to be sold to cover the costs of the 6tb disks.

 

 

Edited March 5, 2022 by thestraycat
adding

[itimpi]

If one is talking about a replacement strategy then there is probably minimal risk on any sensible approach as long as for each stage one keeps the removed (good) disks intact until that stage has successfully completed as you then have the original disks available as your fallback strategy.

[thestraycat]

@itimpi A really great way to summarize. Thanks.

@Frank1940 @JonathanM - Thanks also.

 

If anyone has any experience of replacing more than one disk at a time. I'd love to hear their personal experience.