Hi, last questions please.
let me understand better.
tmpfs is not RAM but HD?
those
tmpfs tmpfs 16G 0 16G 0% /dev/shm
tmpfs tmpfs 16G 33M 16G 1% /run
tmpfs tmpfs 16G 0 16G 0% /sys/fs/cgroup
tmpfs tmpfs 16G 33M 16G 1% /etc/machine-id
are space on HD or max virtual space on RAM ?
just need to add that
tmpfs /tmp tmpfs defaults,noexec,nosuid,noatime,nodiratime,size=512m
(if i want 512M)
to /etc/fstab
(and reboot?)
or something else?
or better to create another partition on disk and then move the /tmp dir?
thanks
On 22/12/2014 19:32, David B Funk wrote:
On Mon, 22 Dec 2014, Reindl Harald wrote:
Am 22.12.2014 um 19:08 schrieb Almond:
so you mean to move the /tmp partition from /dev/md2 to another
partition, since /var/spool/mail and quota control are on the same
partition?
and what size to reserve to the new /tmp... ?
well, we have it on tmpfs like below on our inbound mail-gw and use
in any SA/ClamAV related service (milters and so on) explicit
Environment="TMPDIR=/tmp" to make sure even /var/tmp is not used
also /var/lib/postfix containing nothing else then caches and
lockfiles is on tmpfs to avoid "postscreen_cache.db" become IO bound
and have 99% of all processing in memory, hence also any postfix
configuration is using proxymap
maybe that's why the mailgatewy has a load average of 0.0x all the
time, but keep in mind in our case it is *only* a inbound MX
filtering and scanning messages with no local storage handing over
clean messages to the final destinations via SMTP
tmpfs /tmp tmpfs defaults,noexec,nosuid,noatime,nodiratime,size=512m
tmpfs /var/lib/postfix tmpfs
defaults,noexec,nosuid,noatime,nodiratime,size=64m,uid=89,gid=0,mode=0700
Exactly, but I'd set that /tmp size to something like 10GB or more.
He's got a 2TB root drive so no sense scrimping on /tmp and running
into "fullness" problems later.
