On Mon, Jan 10, 2022 at 2:17 PM Simon <linux@thehobsons.co.uk> wrote:
> Mark Wedel <mwedel@sonic.net> wrote:
>
> > Mythtv workloads are pretty kind on hard drives - it sends to be
> sequential reads and writes, which HDD are pretty good with.
>
> Reads, yes sequential (mostly)
> Writes, definitely not. Each recording causes a minimum of 2 seeks/second
> - write a bit of data, seek, update the metadata, seek, write a bit more of
> the stream, … Especially with multiple recordings, that can be a lot of
> head movement - albeit not as hard work as (say) a well worked database
> without adequate caching.
>
>
>
> > James Abernathy <jfabernathy@gmail.com> wrote:
>
> > When you mirror your SSDs are you using mdadm Mirror or ZFS Mirror. I
> know how to set up a ZFS or mdadm mirrors for data drives but If I use an
> SSD, either SATA or nmve, to boot from the setup for a mirrored boot drive
> is a lot more complicated from the blogs I've seen. How do you do it?
>
> I “just do it” !
>
> I think GRUB has dealt with this for a long time, but also as long as you
> keep to the legacy V1 metadata format for madam mirrored sets, each
> partition in the set will act (read only, for booting from) the same as the
> array itself. That’s because if you use the older format, the metadata is
> at the end of the partition and the actual data (filesystem) starts at the
> beginning of the partition. So you can boot from sda1, sdb1, or the array
> that they are part of - when I first used software raid, I don’t think GRUB
> did handle arrays for booting and “inertia” means that I still tend to use
> the V1 metadata format for my boot partition !
>
>
>
> James Abernathy <jfabernathy@gmail.com> wrote:
>
> > Based on 600 TBW, I could record all the programs in Primetime on the 4
> Major networks in the USA everyday plus some assorted other stuff at an
> average rate of 5GB/hour and use up the SSD in 24.93 years. So if I make
> it to 94, then I'll have to rebuild. :-)
>
> Or it could fail next week - but it’s OK, they’ll send you a nice new but
> blank one to replace it.
> Kingston have just sent me a new 240G drive to replace one that failed. In
> this case I could probably recover the data because it works for a while
> and then “just disappears” off the bus - not that I need to as it was half
> of a mirrored pair.
> Now, if it’s a standalone drive, with years of recordings on it - a
> replacement for the failed drive isn’t much compensation for it’s failure
> :-(
>
>
> Simon
>
I'll probably keep my mythtv and NAS data on mirrored Hard drives and not
SSDs. My current HDs have been working 24/7 without fail for 3 years this
month. I'm a firm believer that most failures of computer hardware
occur during a power cycle. I'll spend the next few weeks testing ZFS,
BTRS, and plane old mdadm on EXT4 to see what I trust for the next system I
build.
Jim A
> Mark Wedel <mwedel@sonic.net> wrote:
>
> > Mythtv workloads are pretty kind on hard drives - it sends to be
> sequential reads and writes, which HDD are pretty good with.
>
> Reads, yes sequential (mostly)
> Writes, definitely not. Each recording causes a minimum of 2 seeks/second
> - write a bit of data, seek, update the metadata, seek, write a bit more of
> the stream, … Especially with multiple recordings, that can be a lot of
> head movement - albeit not as hard work as (say) a well worked database
> without adequate caching.
>
>
>
> > James Abernathy <jfabernathy@gmail.com> wrote:
>
> > When you mirror your SSDs are you using mdadm Mirror or ZFS Mirror. I
> know how to set up a ZFS or mdadm mirrors for data drives but If I use an
> SSD, either SATA or nmve, to boot from the setup for a mirrored boot drive
> is a lot more complicated from the blogs I've seen. How do you do it?
>
> I “just do it” !
>
> I think GRUB has dealt with this for a long time, but also as long as you
> keep to the legacy V1 metadata format for madam mirrored sets, each
> partition in the set will act (read only, for booting from) the same as the
> array itself. That’s because if you use the older format, the metadata is
> at the end of the partition and the actual data (filesystem) starts at the
> beginning of the partition. So you can boot from sda1, sdb1, or the array
> that they are part of - when I first used software raid, I don’t think GRUB
> did handle arrays for booting and “inertia” means that I still tend to use
> the V1 metadata format for my boot partition !
>
>
>
> James Abernathy <jfabernathy@gmail.com> wrote:
>
> > Based on 600 TBW, I could record all the programs in Primetime on the 4
> Major networks in the USA everyday plus some assorted other stuff at an
> average rate of 5GB/hour and use up the SSD in 24.93 years. So if I make
> it to 94, then I'll have to rebuild. :-)
>
> Or it could fail next week - but it’s OK, they’ll send you a nice new but
> blank one to replace it.
> Kingston have just sent me a new 240G drive to replace one that failed. In
> this case I could probably recover the data because it works for a while
> and then “just disappears” off the bus - not that I need to as it was half
> of a mirrored pair.
> Now, if it’s a standalone drive, with years of recordings on it - a
> replacement for the failed drive isn’t much compensation for it’s failure
> :-(
>
>
> Simon
>
I'll probably keep my mythtv and NAS data on mirrored Hard drives and not
SSDs. My current HDs have been working 24/7 without fail for 3 years this
month. I'm a firm believer that most failures of computer hardware
occur during a power cycle. I'll spend the next few weeks testing ZFS,
BTRS, and plane old mdadm on EXT4 to see what I trust for the next system I
build.
Jim A