| 1 |
On Sat, Sep 5, 2015 at 8:21 AM, Nikos Chantziaras <realnc@×××××.com> wrote: |
| 2 |
> On 05/09/15 08:52, Dale wrote: |
| 3 |
>> |
| 4 |
>> tmpfs /var/tmp/portage tmpfs noatime 0 |
| 5 |
>> 0 |
| 6 |
> |
| 7 |
> |
| 8 |
> Off-topic: forget "noatime". Use "lazytime" instead: |
| 9 |
> https://lwn.net/Articles/621046 |
| 10 |
|
| 11 |
I don't really see the point in using either for tmpfs. If the file |
| 12 |
is read its inode is in RAM, and altering it costs just a few CPU |
| 13 |
cycles. On disk it makes a big difference though. |
| 14 |
|
| 15 |
Also, I'd be careful about anything other than noatime on COW |
| 16 |
filesystems like ZFS/btrfs. Disk reads will consume disk space if you |
| 17 |
have any snapshots, since you're modifying inodes. lazytime does't |
| 18 |
really do anything to fix this (I believe the intent is to update |
| 19 |
atime ever 24 hrs as with relatime, but I don't know if this is |
| 20 |
implemented yet). |
| 21 |
|
| 22 |
-- |
| 23 |
Rich |
Archives