memcached timeouts

Nathan Schmidt nschmidt at
Mon May 5 19:47:44 UTC 2008


To reinforce Miguel's point, something we ran into was having two
subnets (separate /27 ip blocks) on the same switch. If you don't
hardcode combined routing for both subnets you'll end up routing some
app-to-memcached traffic upstream to your router and back, which can
cause occasional failures and timeouts depending on the quality of the
upstream connectivity. This was particularly a factor for us because
we have Gig-E on our backplane but our upstream traffic is over a
100mbps PHY.


On Mon, May 5, 2008 at 12:13 PM, Miguel DeAvila
<miguel.j.deavila at> wrote:
> On Monday 05 May 2008 12:05:26 Evert | Rooftop wrote:
>  > I'm getting timeouts trying to connect to memcached. The odd thing is
>  > that the timeouts only occur from 1 single webserver, to 1 single
>  > memcached box:
>  >
>  > Memcache::get() [<a href='function.Memcache-get'>function.Memcache-
>  > get</a>]: Server www5 (tcp 11211) failed with: Connection timed out
>  > (110)
>  >
>  > Other boxes connecting to that memcached instance work fine, and that
>  > webserver connecting to other memcached instances also have no issues.
>  >
>  > This might well be a problem somewhere completely else, but I was
>  > hoping someone on this list can advice on how to track down the exact
>  > problem. What is a good starting point?
>  - ping from the affected web server to the affected memcache server
>  - nc from the affected web server to the affected memcache server
>  - compare output of /sbin/route on affected web server with normally-functioning web server

New office! 1825 South Grant Street, Suite 850 San Mateo 94402
New home! 21677 Rainbow Drive Cupertino 95014
New phone! 415.420.1647

More information about the memcached mailing list