The cost of atime is not very dramatic when using zfs except for
perhaps the snapshots issue which was already mentioned. This is
because the atime updates are cached in memory and are only written to
underlying store at the next zfs sync interval. So maybe atime will
be updated in the filesy
Seems like the Linux style O_NOATIME flag (usable only by root - or on Solaris
derivatives, possibly with a fine-grained permission) might be a useful
addition to the OS, so that system programs that read a lot of files can avoid
causing massive time updates that either generate excess I/O and o
On Sat, Aug 21, 2021 at 10:32 AM James wrote:
> On 19/08/2021 23:58, Carl Brewer wrote:
> > Further to this - is it worth disabling atime on the ZFS root pool
> > that's on the SSDs? I don't imagine it's a lot of data, but it would add
> > up over the years.
>
> Do you ever look at the access tim
On 21/08/2021 7:32 pm, James wrote:
On 19/08/2021 23:58, Carl Brewer wrote:
Further to this - is it worth disabling atime on the ZFS root pool
that's on the SSDs? I don't imagine it's a lot of data, but it would add
up over the years.
Do you ever look at the access times?
not personally, bu
On 19/08/2021 23:58, Carl Brewer wrote:
Further to this - is it worth disabling atime on the ZFS root pool
that's on the SSDs? I don't imagine it's a lot of data, but it would add
up over the years.
Do you ever look at the access times?
SSD or otherwise it must add writes when reading.
I have b
