[mythtv-users] Can't get storage priorities correctly.

[Martin Compton]

On 11/18/23 14:14, Ram Ramesh wrote:
> On 11/18/23 04:17, Stephen Worthington wrote:
>> On Fri, 17 Nov 2023 22:06:02 -0600, you wrote:
>>
>>> Thanks for your concern. I thought a  lot before choosing this path.
>>>
>>> Yes, I did my calculations. My SSDs have 600TBW and at 30GB per day, I
>>> am looking at 54 years of life (600*1000/30/365 ~ 54). I also have
>>> smartd doing its thing. It will report when SSDs start using their spare
>>> blocks.  At about $60/TB (for 54 years of claimed life), I feel it is a
>>> fair game.
>> That certainly sounds good.  But do check the smartctl -a output - my
>> new 1 Tbyte SSD in my main MythTV box is only 214 days old, and is
>> currently showing 16.1 TBytes written without me doing any recording
>> to it (75 Gbytes per day).  It says it is 1% used, so it is still
>> going to last a long time, even at that rate.
>>
>>> So far, in my mythtv usage (10+ years), I have had 10+ spinning disks
>>> die. I have still my very first SSD (256G SATA) and several others  as I
>>> have no use for them with newer/larger  SSDs replacing the old ones.
>>> So, I am looking to try them out to see if TBW values really happen.
>> Over the last 10 years I think I will have had 10+ hard drives die
>> also, but only 2 of them prematurely.  The others had been running for
>> a long time, over their expected lifetime.  I now am buying enterprise
>> class drives (20+ Tbytes) with 5 year warranties, but they have not
>> been around for more than 5 years yet.  I do have two ancient
>> enterprise class 3 Tbyte HGST drives that I swapped out because they
>> were too small after 9-10 years 24/7.  I put one of them into my
>> mother's MythTV box running 24/7 again a couple of months later.  It
>> is currently showing 105553 power on hours (12 years) and is showing
>> no problems at all.  They were not cheap when I bought them, but they
>> do show that enterprise class drives can last a very long time.
>>
>>> Suppose I am wrong and my SSDs show wear, it is a single mythtv-setup
>>> change that will get back to trashing spinning disks.
>> Yes, very easy to do.  I really like how storage groups work.
>>
>>> When I record and watch at the same time, the amount of seek is very
>>> noticeable (especially with RAIDs) and I prefer the random access
>>> capabilities of SSDs.
>> With my enterprise class drives, I can have them recording 2
>> programmes at once and playing back at the same time with no problems.
>> >From the specifications, I would think three recordings at once would
>> be fine too, but I never get to that as I have 7 recording drives.  I
>> can certainly hear them seeking, but it does not sound over the top at
>> all.  But I can also do 2 recordings plus playback from my old (very
>> quiet) WD Green 4 Tbyte drives, which have much lower specifications.
>> They are currently showing 81029 and 77591 power on hours (9.2 and 8.9
>> years).  So even cheap drives can last well if you get lucky and they
>> were over designed.
>>
>> I am wondering if RAID 1 is the problem somehow.  If the data was
>> being written to the first drive and then read back again to be copied
>> to the second drive, that might cause what you are getting.  But that
>> would be a really bad way of doing RAID 1 when you can just write to
>> both drives at once from the same buffers.
>>
>>> Regards
>>> Ramesh
>> _______________________________________________
>> mythtv-users mailing list
>> mythtv-users at mythtv.org
>> http://lists.mythtv.org/mailman/listinfo/mythtv-users
>> http://wiki.mythtv.org/Mailing_List_etiquette
>> MythTV Forums:https://forum.mythtv.org
> What I meant by noticeable is that I can hear the disks and see the 
> activity lights and disks hunting. No issues with actual myth 
> performance. I simply think that the seek should be avoided, if 
> possible. It just annoys me and makes me wonder if the wear-and-tear 
> is eventually going to get my disks.
>
> Yes, my 2TB HGST (enterprise) is running past its 10th year. BTW, I 
> bought this refurbished from an unknown vendor on ebay. Clearly they 
> used to make good enterprise disks. Now everything is really spinning 
> rust.  I only buy enterprise spinners and that too across brands and 
> batches so that I really get average life from them. However, I am not 
> confident on spinners anymore and they are also very big these days 
> making any operation risky of in process failures. I will learn in the 
> next 10 years if SSDs are just a different kind of rust or they 
> actually live long like other electronic circuits. Early experiments 
> on SSDs do indicate that they could be the latter.
>
> Regards
> Ramesh

I've got several hard drives with about 9 years on them.  Two (out of 6) 
failed, but did so in a way that I was able to move most recordings off 
of the failing drive.

Definitely worth the extra money to get an enterprise/nas drive. Mine 
have failed gracefully in the last 5 years or so.  The newer larger 
(12TB+) drives make me a little nervous, but not enough to implement RAID.

I was a little worried about thrashing hard drives, so I put lots of RAM 
(32GB) on the backend/file server.  That way it can run commercial skip 
on an active recording without reading from the actual disk as the file 
data still exists in the operating system's file buffers.

I also have write caching enabled since the server is on a UPS (highly 
recommended).  At one time I would have up 3 simultaneous recordings 
using one hard disk with pretty smooth disk usage.

I have the mythtv database on an SSD, and it does not seem to wear much, 
according to smartctl only about 6.5TB written and it is almost 2 years old.

It will be interesting to see how SSD technology wears down.  I have 
only had a few and none have failed yet.