On 11/24/2014 07:55 AM, Gordan Bobic wrote:
On 11/24/2014 12:47 PM, Robert Moskowitz wrote:
It isn't exactly lightweight. That's why I am sticking with only
nolisting and clam and uri milters without spamassassin.
It might also be worth checking what your disk I/O and swapping are
like, since on a Pi clamd alone will be eating about 1/2 of the RAM.
# free
total used free shared buffers cached
Mem: 2050624 1584976 465648 360 265588 761744
-/+ buffers/cache: 557644 1492980
Swap: 4193276 1276 4192000
How can I get a measure of disk I/O?
iostat -x 1
varying numbers. Here is one of the higher ones:
avg-cpu: %user %nice %system %iowait %steal %idle
3.21 0.06 3.25 0.38 0.00 93.11
Device: rrqm/s wrqm/s r/s w/s rsec/s wsec/s avgrq-sz
avgqu-sz await svctm %util
sda 0.00 0.32 0.01 0.49 0.45 13.22 27.20
0.02 31.05 17.69 0.89
avg-cpu: %user %nice %system %iowait %steal %idle
45.00 0.00 6.00 3.00 0.00 46.00
Device: rrqm/s wrqm/s r/s w/s rsec/s wsec/s avgrq-sz
avgqu-sz await svctm %util
sda 0.00 25.00 0.00 4.00 0.00 232.00
58.00 0.06 15.00 15.00 6.00
Is this in a random period when you were watching or during a period
when you were experiencing overloads? You really should set up some
kind of monitoring (nagios/cacti/zabbix) and get some high resulution
graphs on what's going on with various aspects of load (cpu usage,
load average, disk I/O) at the specific times when the overloads are
occurring.
I am going to have to select and learn how to use one of these tools.
If you are seeing the %util abouve 80% you are probably bottlenecked
on disk I/O.
iotop might also help.
Had to install this, and it really does not report any activity. Things
are probably happening too fast for too short of a time.
I believe it show cumulative figures between refreshes, like top.
And they were showing zeros. Once in a while some number, but can't get
a capture of them.
But the tables are small. domain has 4 entries and is hit all the
time. mailbox has a dozen entries and is hit for every incoming message
at least once. Does mysql do a good job of caching?
Yes, provided it's configuration is appropriate (big enough buffer
pools and table handle caches).
Going to have to read up on what the defaults are, how to measure, and
how to tune.
_______________________________________________
users mailing list
[email protected]
http://lists.redsleeve.org/mailman/listinfo/users
