| 1 |
On 30/01/2017 01:06, Mick wrote: |
| 2 |
> On Sunday 29 Jan 2017 22:10:59 Neil Bothwick wrote: |
| 3 |
>> On Sun, 29 Jan 2017 19:17:47 +0000, Mick wrote: |
| 4 |
>>>> Are you running fstrim once in a while like it's recommended? |
| 5 |
>>>> Apparently using 'discard' as an option when mounting is no longer |
| 6 |
>>>> recommended. On my laptops I use a systemd timer to do this. Before |
| 7 |
>>>> that I used anacron (I think it was anacron) which would run missed |
| 8 |
>>>> cronjobs. |
| 9 |
>>> |
| 10 |
>>> This is surprised me ... I just installed Gentoo on a MacBook and the |
| 11 |
>>> handbook/wiki said to use discard in fstab ... I'm running two PCs like |
| 12 |
>>> this now. :-/ |
| 13 |
>>> |
| 14 |
>>> Is there a URL somewhere recommending otherwise? |
| 15 |
>> |
| 16 |
>> man fstrim: |
| 17 |
>> |
| 18 |
>> Running fstrim frequently, or even using mount -o discard, might |
| 19 |
>> negatively affect the lifetime of poor-quality SSD devices. For most |
| 20 |
>> desktop and server systems a sufficient trimming frequency is once a |
| 21 |
>> week. Note that not all devices support a queued trim, so each trim |
| 22 |
>> command incurs a performance penalty on whatever else might be trying to |
| 23 |
>> use the disk at the time. |
| 24 |
> |
| 25 |
> Hmm .... I better take these discards off fstab then. Are these weekly trims |
| 26 |
> OK, if the PC is rebooted on a daily basis? |
| 27 |
> |
| 28 |
|
| 29 |
You can deal with trim as if it were a low-impact defrag on Windows. All |
| 30 |
trim really does is clean up blocks from deleted files, nuke the |
| 31 |
metadata and return the blocks to the unallocated pool. |
| 32 |
|
| 33 |
This is an expensive operation on SSDs which is why they are delayed. In |
| 34 |
theory you could even leave the disk untrim'med until you need the space :-) |
| 35 |
|
| 36 |
|
| 37 |
-- |
| 38 |
Alan McKinnon |
| 39 |
alan.mckinnon@×××××.com |
Archives