Strange memcached behavior on stress test

[Jamie McCarthy]

sgrimm at facebook.com (Steven Grimm) writes:

>  > We restart it daily.
> 
> We, on the other hand, never restart it, and we transfer way more
> than 64GB of data with no signs of instability. I just looked at
> one of our memcached instances and it has these stats (commas
> added for readability):
> 
> STAT bytes_read 204,016,350,632
> STAT bytes_written 11,728,604,145,226
[...]
> This is after about 30 days of uptime.

I was curious, so I checked one of Slashdot's.  Like most of our 16
memcached servers, this one has been running for five months since
its last restart:

STAT uptime 13,052,212
STAT total_connections 1,690,123
STAT cmd_get 472,869,567
STAT cmd_set 23,020,413
STAT bytes_read 87,552,614,529
STAT bytes_written 830,072,934,139

Its CPU usage according to ps has averaged around 0.91%.

We only restart them when code changes would result in wrong/stale
data or when the box itself gets rebooted.
-- 
  Jamie McCarthy
 http://mccarthy.vg/
  jamie at mccarthy.vg