Re: [SQU] [From CVS] cache no longer working ...

Okay, not sure who maintains the FAQ, but this might be great for it, as I
did check there first :)

According to src/fs/diskd/store_diskd.h, SHMBUFS is 96, and they recommend
Q2 (MAGIC2??) to be no more then 75% of that ... now, according to that
file:

/*
 * SHMBUFS is the number of shared memory buffers to allocate for
 * Each SwapDir.
 */

so, if I have three swapdirs, then I need 96*3 is the numver of shared
memory buffers I need to have configurd in my kernel, right? based on the
FAQ, I have:

options SYSVSHM #SYSV-style shared memory
options SHMSEG=16 # max shared mem id's per process
options SHMMNI=32 # max shared mem id's per system
options SHMMAX=2097152 # max shared memory segment size (bytes)
options SHMALL=4096 # max amount of shared memory (pages)
options SYSVMSG #SYSV-style message queues
options MSGMNB=16384 # max # of bytes in a queue
options MSGMNI=41 # number of message queue identifiers
options MSGSEG=2049 # number of message segments
options MSGSSZ=64 # size of a message segment
options MSGTQL=512 # max messages in system
options SYSVSEM #SYSV-style semaphores

right now (FreeBSD system) ... I've never been able to adequately figure
this out ... SHMBUFS == ?? SHMSEG? Looking at the above, I'm suspecting
that I don't have enough allocated at the kernel level, but am not sure
which one I should be looking at :(

On Tue, 21 Nov 2000, Adrian Chadd wrote:

> On Tue, Nov 21, 2000, Marc G. Fournier wrote:
> >
> > Okay, next the stupid question ... I'm using diskd ... what shmbufs are
> > you referring to? Are you referring ot the Q1/Q2 parameters in
> > squid.conf?
> >
> > cache_dir diskd -1 /usr/local/squid/cache/1 15240 16 256 64 72
> > cache_dir diskd -1 /usr/local/squid/cache/2 15240 16 256 64 72
> > cache_dir diskd -1 /usr/local/squid/cache/3 15240 16 256 64 72
> >
> > Any recommendations on what to set them at?
>
> Yup. You should look at upping those two values until you hit a hitrate
> which is acceptable under your disk load.
>
> Unfortunately if you just up them you'll find mysterious crashes as
> you run out of sysvshm. src/fs/diskd/store_diskd.h defines SHMBUFS
> which is the actual queue size (whereas the q1/q2 determine when
> scheduleable disk ops and forced flushing occur..) .. you will need
> to up that to be slightly bigger than q2.
>
> And yes, it'll be tidied up before squid-2.4 is released. :-)
>
>
> Adrian
>
> --
> To unsubscribe, see http://www.squid-cache.org/mailing-lists.html
>
>

Marc G. Fournier marc.fournier@acadiau.ca
Senior Systems Administrator Acadia University

  "These are my opinions, which are not necessarily shared by my employer"

--
To unsubscribe, see http://www.squid-cache.org/mailing-lists.html