Hi Duane
> >I am having a couple of problems with squid 1.1.1:
> >I have been running this version for a few weeks without hassles on
> >a linux machine, kernel 2.0.24
> >
> This looks perfectly normal. The four select() calls on FD 20 are
> checking the ICP socket. Notice after that there is a write() on
> FD 843.
hmm - strange.
> My guess is that your cache stopped accepting connections because
> it was close to the max number of descriptors. Squid waits for
> current connections to go away before accepting more.
Something was strange here:
356 CLOSE
212 CLOSE_WAIT
493 ESTABLISHED
36 FIN_WAIT1
1 TIME_WAIT
There is a limit of 1024 filehandles (less about, say 7 for the logs etc)
Do "CLOSE" or "CLOSE_WAIT" sockets still consume filehandles?
There is no way to know without getting a signal of some kind to dump the
stuff that the cache manager interface returns.
The other option is to dedicate a different port to cache managers, with a
dedicated filehandle, so that the cache manager will always be able to
connect.
Which do you think is better?
> >Cache2 runs 1.1.1
> >newt:~/squid/echoping> ./echoping -n 5 -h http://www.vodacom.co.za/ cache2:808
> 0
> >Elapsed time: 3.242044 seconds
> >Elapsed time: 3.804418 seconds
> >Elapsed time: 3.010914 seconds
> >Elapsed time: 3.101601 seconds
> >Elapsed time: 3.144090 seconds
> >---
> >Minimum time: 3.010914 seconds (85 bytes per sec.)
> >Maximum time: 3.804418 seconds (67 bytes per sec.)
> >Average time: 3.260613 seconds (79 bytes per sec.)
> >Median time: 3.144090 seconds (81 bytes per sec.)
> >newt:~/squid/echoping>
> >
> >Clearly squid 1.1 is getting the page from the real site every time, with
> >the logs showing
> >852973763.314 3115 196.23.0.84 TCP_REFRESH_MISS/200 4716 GET http://www.voda
> com.co.za/ - DIRECT/www.vodacom.co.za text/html
> >
> >But if I look at the headers there doesn't seem to be any "meta" info... and
> >cache manager shows:
> >17.http://www.vodacom.co.za/
> > Verified 27Y ago, Used 3s ago, Modified 19s ago, Expires 27Y ago,
> > 0 bytes, 1 accesses, 1 active clients,
> > STORE_PENDING, NOT_IN_MEMORY, NO_SWAP, PING_NONE,
> > 2 Locks, Flags: EC,KC,HT,ED,KP,HI
>
> There is no last-modified time. For these Squid usually issues
> IMS requests because it can't calculate the 'age factor'.
> Doing so would result in a divide-by-zero.
In squid 1.0 you could give it a default time for pages with no meta info,
so pages with meta info were treated as a special case (which they are,
considering the number of pages that don't have them.
Say I wanted to:
Cache all pages without meta info for 3 days
cache pages with meta info for 40% of the time they are "fresh" for, but not
less than 1/2 an hour, and not longer that 3 weeks
Can I no longer do this?
Oskar
=============================================
'experience made art, but inexperience luck.'
=============================================
Received on Tue Jan 14 1997 - 05:07:20 MST
This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:34:04 MST