Re: cvs commit: squid/src stat.c

Adrian Chadd wrote:
>
> On Tue, Oct 03, 2000, wessels@squid-cache.org wrote:
> > wessels 2000/10/03 14:51:04 MDT
> >
> > Modified files:
> > src stat.c
> > Log:
> > DW:
> > - This ugly hack is here to prevent the user from seeing a negative
> > byte hit ratio. When we fetch a cache digest from a neighbor,
> > it gets treated like a cache miss because the object is consumed
> > internally. Thus, we subtract cache digest bytes out before
> > calculating the byte hit ratio.
> >
> > Revision Changes Path
> > 1.335 +16 -1 squid/src/stat.c
> >
>
> Would this explain why you see negative byte ratios on non-active caches
> due to cachemgr activity? They're also treated as TCP_MISS ..

cachemgr access most likely count in the byte hit ratio statistics yes,
but probably as hits.. Byte hit ratio is calculated as the difference
between retreived data and data sent to the clients
(server.all.kbytes_in related to client_http.kbytes_out)

Regarding the digest traffic. In my opinion retreival of digests should
count as misses and affect the byte hit ratio. After all Squid is
fetching content remotely. If you want to get fancy then add a "fast"
ACL to deny certain requests from the statistics, but hacking digest
retreival to alway be excluded from the statistics is wrong I think.

Digest providing and other internal objects should be excluded as these
are handled locally, but are not actually cache hits...
Received on Wed Oct 04 2000 - 01:34:35 MDT