> > Few months ago i got one patch from you to let the linear raid handle >2TB > > devices. > > At this point i not to able to test it, because i dont have money to buy > > the upgrade. > > > > The question is this: > > > > If i switch from i386 to x86_64, the patch will be unneccesary or still need > > the kernel, to handle the big drive? > > You need it on x86_64 as well, though it you are running 2.6.14 or > later, the patch is already there. > > NeilBrown Hello, Neil, Now it is the time to upgrade my system from 8TB to 13.2TB. :-) (Simple reminder if somebody forget, or missed something: i use 4 disk node with NBD. Each nodes are 2TB (11x200GB Raid5) now, and in the concentrator i use one raid0 to join the nodes to one big raid array. Some months ago, when i try to build this system, found one limit. The linear array cannot import the 8TB array, caused by its size.) I want to ask you, Neil, about possible limitations.... My plan: I will replace the all 200 GB hdd with new 300GB, + add some new (12.) drive on each node. + i want to backup the existing data to the new space. :-) (or backup some data if i unable to backup the all of 8TB) Step 1: fdisk. On the new disks, i will create one 200GB partition(part.2.) on the *END* of the devices (only on 11 device) similar with the original size of the existing 200GB hdds... And another one from the beginning, to the next partition.(part.1. ~ 100GB) (I have found the first problem: [called as #1] The original 200GB devices partiton starts from the beginning of the device, and this is only from head 1, and not head 0, caused by MBR.) Step 2: I will backup the original nodes to this new raid5 on the end of the devices. old md0 = 2TB-64k, new md0 = 2TB There is no problem. :-) Step 3: Join the 4 new 2TB(4x11x200GB) device on the concentrator as md1 (inside the existing 8TB data), and the another 4 new device(4x12x100GB) as md0 (empty array). There is some problem, caused by #1. The new 4x2TB devices is bigger with +64KB, and the raid0's superblock is "wrong placed". This is OK, i will use mdadm --build /dev/md1 command to buid an array, without the superblock. :-) Step 4: Copy the most valuable data from md1 to md0 The md1 > md0. This is OK, the problem is simly mine. :-D (I need to delete....) Step 5a: I will delete the partition 2 from all HDD, and resize the partition 1 to fit the whole drive (300GB). With this, the nodes will be 3.3TB! (12x300 [-1 raid5]) Step 5b: Resizing the md0 (raid5 )in the nodes. If i resize the partitions, the superblock is "wrong placed" again. It is safe to re-create the raid5 array, with the new size? In my previos mail with title"where is the spare :-)" i can see, there is some risk too! :-/ If the newly created raid5 array starts again to using the 12. drive as spare, it will owerwrite everything!!! Step 6: I will recreate the md0 on the concentrator, with the new node size. (4x3.3TB) Step 7: Finally resize(8TB -> 13.2TB) the XFS on the md0 in concentrator. Done. I hope, you can say something else than "good luck!" :-D Thanks, Janos (Happy new year! :-) - To unsubscribe from this list: send the line "unsubscribe linux-raid" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
