Currently in a horrible situation, disk has failed so in a degraded state. However all the drives were purchased at the same time (in the same order). So worried about a second drive failing and losing the data, currently have two separate arrays (mhddfs) for general stores of data (one monthly backup of the zfs and just general storage) both have enough space to copy to the data.
So my question is to rysnc my data to another drive, or wait until I have a replacement drive and just not use the server and then rysnc but risk the resilver. Just curious on ideas on what might be the safest thing to do.
What disks do you have. perferably model number, but we can work with size, make and series. Depending on the disk, we need to suggest different solutions.
What OS are you using?
EDIT: also, how much data (approx) do you have? If you've only got 50-350ish GB, you should be okay doing either.
Another option you have is to take an image of all the drives and mount the images.
More of a general question of what u would do in my situation rather than anything else. Rsync is rsync and resilvers are gonna take pretty much the same on any system (but using debian atm). But if you think it will help sadly my arrays are getting quite large
About 6tb space on each array, with the ZFS array nearly full at 5.7 tb so i'm pretty screwed there, as either way, times are gonna be long. My scrub takes a while (think it's about 9 hrs but I just leave it overnight most times), Hard disks listed below.
3tb x 3 WD Reds 2tb x 3 WD Reds and 6 tb of junk drives (3 one large 2 small)
It's more of a question of what will put the least amount of stress on the drives over time, all i care about is the data really and know what needs to be done just hesitant to start before I got others opinions.
With the reds, I'd feel safer rsyncing a copy of the data, starting with the most important stuff. Once you have it all copied off, you can decide if you want to rebuild the array or replace all the drives. Since you mentioned you've got it nearly full, I would seriously consider a 3x6TB hgst array. (I've had good results with HGST in nas applications)
You're going to wind up putting less stress on the drives rsyncing the data away.
How old are the drives? This is surprising to see these drives die so quickly. This makes me want to switch to raidz2.
Keep in mind you may be able to get this failed drive warrantied by WD.
Yea would love to buy HGST drives but sadly it's a nightmare to find anyone who sells them in the UK which doesn't help (otherwise I probably wouldn't have gotten reds).
Yea what I thought, as thought if anything would kill the other drives its the excessive writes needed for the resilver.
Bit disappointed on that front as they are only about 2 years old especially when you realise I have drives that are 10 years old that are still working (hell even my green WD drive has lasted 5 years). Kind of in agreement with you however wondering if I should just get another drive and use raidz1 with a extra hotswap or raidz2, but having an extra drive in the array sounds like the better option ironically I'll probably end up with raidz2 and a hotswap just incase :P.
Yea having the fun of finding the dead drive (wish I marked them), then got that fun later hopefully though should be able to get a replacement as it's under the 3 years warranty.
I also prefer to run smartctl; should be part of the smartmontools package on debian.
smartctl -a /dev/xyz should give you all the information you need to determine which drive is bad, and also I would check all the other drives as well; this should give you possibly some warning signs of any other drive(s) possibly failing.
You might want to look at zfs send / zfs receive. They can be used to send and receive zfs snaphots to another zfs pool or you can put the output of zfs send into compression and compress it to a backup file anywhere on your system that has enough space.
Well that was kind of painless, can't remenber what program it was but just pulled off the serial number. All i will say is WD replacements are amazing, got a replacement in 2 days and then sent off my drive back to them (it's called advanced rma). Worked great and was very painless, and for a bonus they replaced it with a Red Pro, so as you can imagine was delighted. Seems it was just one duff drive as all the rest are purring along nicely.
Cheers the the help guys oh btw fell in love with blkid so handy for grabbing UUID's :).