So, I am creating a new file server for large files (movies etc, over 100MB files), and am wondering what FS to use. I will be running RAID 5 or 6, probably on a 3Ware 9650SE-12ML card, with several Seagate 1Tb disks. OS will be CentOS 5.2.
I was thinking XFS would be the best choice of filesystem (even though the XFS modules are not included by default in this OS), however it seems that XFS on LVM2 (which is what I would be implementing) has issues if 4K stacks are enabled, as happens by default, especially in 2.6 kernel Red Hat distros. The issue being that under high I/O, the system locks up.. Obviously this is not an ideal situation, so what I would like to know is:
Has anyone experienced this? What is the likelihood of such an event happening? Am I being overly paranoid here?
Also, XFS on linux does not allow block sizes greater than the page size, which is 4K by default. This seems a pointlessly small block size for a file system that does not need to deal with small files, and not all that many directories (pretty flat file structure). Especially when you consider that I will be looking at a RAID 6 chunk size of 1Mb, or maybe 512K (the bigger the better as far as I can tell). How hard is it to increase the pagesize for the kernel, to allow bigger block sizes? Or am I making this a whole lot more complicated than I need to?
Next question:
how does one create a stride aligned XFS filesystem? http://linux-raid.osdl.org/index.php/RAID_setup#Options_for_mke2fs thats great for ext2, but surely we can do something similar for XFS? Have been reading on linux.com: http://www.linux.com/feature/140734 that this is a good thing to do, but there isn't much info that I can see there about how to do it.
Any other advice you would give for this?
I was thinking XFS would be the best choice of filesystem (even though the XFS modules are not included by default in this OS), however it seems that XFS on LVM2 (which is what I would be implementing) has issues if 4K stacks are enabled, as happens by default, especially in 2.6 kernel Red Hat distros. The issue being that under high I/O, the system locks up.. Obviously this is not an ideal situation, so what I would like to know is:
Has anyone experienced this? What is the likelihood of such an event happening? Am I being overly paranoid here?
Also, XFS on linux does not allow block sizes greater than the page size, which is 4K by default. This seems a pointlessly small block size for a file system that does not need to deal with small files, and not all that many directories (pretty flat file structure). Especially when you consider that I will be looking at a RAID 6 chunk size of 1Mb, or maybe 512K (the bigger the better as far as I can tell). How hard is it to increase the pagesize for the kernel, to allow bigger block sizes? Or am I making this a whole lot more complicated than I need to?
Next question:
how does one create a stride aligned XFS filesystem? http://linux-raid.osdl.org/index.php/RAID_setup#Options_for_mke2fs thats great for ext2, but surely we can do something similar for XFS? Have been reading on linux.com: http://www.linux.com/feature/140734 that this is a good thing to do, but there isn't much info that I can see there about how to do it.
Any other advice you would give for this?
Facebook
Twitter