Re: [squid-users] Re: does rock type deny being dedicated to specific process ??

On 27/10/2013 8:01 p.m., Ahmad wrote:
> hi amos ,
> i read bad news about rock when rock shared between process , i read that it
> reduce hit ratio !
>
> i read form
> http://wiki.squid-cache.org/Features/RockStore
> it says :
> /Objects larger than 32,000 bytes cannot be cached when cache_dirs are
> shared among workers. Rock Store itself supports arbitrary slot sizes, but
> disker processes use IPC I/O (rather than Blocking I/O) which relies on
> shared memory pages, which are currently hard-coded to be 32KB in size. You
> can manually raise the shared page size to 64KB or even more by modifying
> Ipc::Mem::PageSize(), but you will waste more RAM by doing so. To
> efficiently support shared caching of larger objects, we need to teach Rock
> Store to read and write slots in chunks smaller than the slot size. /
>
> as i understood , the max object size for disk caching will be 32 k ,
> am i correct ???
> i see it will be slow writing on hardisks an slow caching !
>
> am i correct ???

*for things stored in the rock cache_dir* only. non-SMP cache_dir such
as AUFS still cache larger items.

Im not sure if the imitation applied to memory cached objects, but when
SMP is enabled that is likely as well.

If this is a critical issue for you please try out the large-rock
experimental feature branch. It has changes which remove those
limitations and also includes collapsed-forwarding port from squid-2.7
to allow backend fetches to be HIT on before they have finished arriving.

> thats why i want to give each hardisk of my disks of rock type a single
> process !!

The limitation applies to rock storage type regardless of SMP sharing.
It is designed to work with in those same limits.

> =================
> also , im not understating u here ,
> * If you use ${process_number} or
> ${process_name} macros these channels are never setup and things WILL break.
> *
>
> ????

With three workers and a rock cache there are actually 5 processes running:

kid1 - worker #1 ... ${process_number} = 1
kid2 - worker #2 ... ${process_number} = 2
kid3 - worker #3 ... ${process_number} = 3
kid4 - rock disker #1 ... ${process_number} = 4
kid5 - coordinator ... ${process_number} = 5

[I'm not completely sure of remembering order between coordinator and
disker, may be the other way around].

If you configure each process to access a different FS directory name
for the rock dir. You end up with disker creating a rock DB at /rock4
when the backend workers trying to use /rock2, and /rock3. The
coordinator thinks the rock dir exists at /rock5.
* None of the processes will accept SMP packets about altering or
fetching rock dir contents in an area they are not configured to use.
* the workers will try to connect to diskers setup for the /rock2 and
/rock3 - which do not exist. This is the shm_open connection error you see.
* the other /rock4 message is disker or coordinator trying to open to
receive messages.

Amos
Received on Sun Oct 27 2013 - 08:20:38 MDT