sgrimm at facebook.com
Wed Aug 9 19:07:20 UTC 2006
Ah, yes, excellent point! If you're graphing the total size of items in
the cache, rather than the process memory usage, then 700MB is about
right for 1.1.12. Grab 1.1.13pre2 or search this list for my memory
efficiency patches to 1.1.12 if you want to increase that; with my
patches the efficiency is more like 85-90%.
When you're at 700MB, what does "top" show for the total process size?
Matthew Nielsen wrote:
> I don't know how cacti shows its graphic on this (amount used by the
> total process versus the memory used by the items in the cache), but
> it's worth noting that memcache is only about 70% efficient at the
> moment so getting 700MB of usable cache out of -m 1024 is inline with
> what we'd expect.
> -Matthew Nielsen
> On Aug 9, 2006, at 12:54 PM, Brett Richter wrote:
>> thanks for the quick reply... the line def says -m 1024 but it never
>> goes higher than 700mb (we use cacti graphs). And yes those werid
>> debian guys! it uses a conf file..
>> Any other suggestions?
>> On Aug 9, 2006, at 2:48 PM, Steven Grimm wrote:
>>> Memcached doesn't have a config file per se (must be something the
>>> Debian guys added) so the first thing to look at is what command
>>> line parameters are getting passed to the actual memcached process.
>>> The one you want to look at is "-m", which controls the cache size.
>>> If you run "memcached -m 1024" then the cache should grow up to 1GB.
>>> As for using all three of your servers, that is most likely a client
>>> configuration issue. The client is responsible for choosing which
>>> server to use for a particular key; if the client is only configured
>>> to know about one of the servers, it will never contact the other two.
>>> Brett Richter wrote:
>>>> I'm sure this has been asked a million times but we seem to be
>>>> running up against a memory ceiling on memcached. We use it to
>>>> cache dynamic content of our website. No matter how high I set the
>>>> memory limit in the conf file it never goes higher than 700mb (give
>>>> or take). We definitely need more than that as any hit that
>>>> doesn't hit the cache is going to the database and its getting
>>>> overwhelmed at times.
>>>> A bit about our setup. We have 3 servers running memcached, both
>>>> set to 1024mb memory. It seems that it's really only using 1
>>>> server right now (looking into why its not using the pool of 3), so
>>>> I really need to get one up to like 3gb till we can look at the
>>>> code and see why its not using the memcache pool better (or at all).
>>>> Systems are Debian Stable 64 using one of their packaged memcached
>>>> installs version 1.1.12-1 of memcache
>>>> Sorry if this seems cryptic of not enough detail I'm the sysadmin
>>>> not the developer of the back end content manager.
>>>> Any help would be great!
More information about the memcached