Re: incomplete loads

I see that the error regarding "File_Map_Allocate appears to have occurred fairly
recently going back through the last couple of months postings, and it would appear
that this is possibly a related problem .

I started getting the message recently regarding "File_MAp_Allocate" .
The site is running an early version of squid (ver 1.1.22) on a Pentium I (133) m/board
with 64 meg. using BSD 2.2.8 and until the last couple of months it hadn't caused any
grief whatsoever. It would appear from the study of the recent mailing lists which have
reported this error that this is also happening on more recent versions so I doubt that it
is version related. In addition it occurs, I've read, under Red Hat and SunOS so
presumably it is NOT UNIX operating system flavour specific either.

With a restart, or reboot even, squid appeared to start and run OK, but the problem
re-occured when someone on a Windows box attempted to access a site. Most of the
html page text would come down but it would get stuck on the graphics image files
(jpeg or gif).

I have yet to read a reply and I could find no reference in the FAQ or docs, but I solved
the error by reducing the store_avg_object_size from 13k to 8k (where the cache_mgr
settings are) in the config file and restarting. The PC has behaved itself for 3 weeks
now.

I found a message in the logs which had appeared as a warning notice several days
earlier (prior to SQUID finally fully objecting to any further seamless operation) advising
to lower this value as the cache was utilizing 88%. I took a guess at the value of 8 which
I figured would reduce the 88% to approx. 60% (somewhere between half-full and
two-thirds full). Well it's currently operational and has ceased to complain anyway
which is the bottom line.
 
I had NOT heeded the said warning soon enough.

Moral: Pays to read your log files judiciously.

Perhaps Henrik you can enlighten us with more info on this particular variable.

G. Anderson

PS. Apologies if it's already in the FAQ but I didn't find it if it is.

On Fri, 22 Oct 1999 01:56:31 +0200, Henrik Nordstrom wrote:

>Tilman Schmidt wrote:
>
>> and I am rather frequently seeing files being incompletely loaded - ie. HTML
>> pages breaking of in mid-text or images showing only the top part.
>
>That's not nice at all.
>
>> A simple browser reload only shows the same incomplete info again, but
shift/reload
>> in most cases gets the complete info.
>
>Quite normal on such errors regarless of how they are caused. The
>browser incorrecly thinks it has the whole page, and a plain reload only
>verifies that the page hasn't been modified which it hasn't.
>
>> 940509416.770 41246 HeartOfGold.sema.de TCP_MISS/200 27312 GET
http://news.cis.dfn.de/de/hierarchies.html -
FIRST_PARENT_MISS/www-cache.xlink.net text/html
>> 940511646.728 133775 HeartOfGold.sema.de TCP_MISS/200 27203 GET
http://news.cis.dfn.de/de/hierarchies.html - PARENT_HIT/www-cache.xlink.net text/html
>
>Hmm.. it looks like you got the whole page on both requests. The odd
>thing is that the reply of the first (failing) request was 109 bytes
>larger than the second (not failing)..
>
>It would be interesting to see the store.log entries for these requests.
>
>> The portion of cache.log around that time shows:
>>
>> 1999/10/21 14:17:17| Detected DEAD Parent: www-cache.xlink.net/3128/3130
>> 1999/10/21 14:17:18| Detected REVIVED Parent: www-cache.xlink.net/3128/3130
>> 1999/10/21 14:28:01| TCP connection to www-cache.xlink.net/3128 failed
>
>Your parent seems to have servere performance problems. It is not
>answering ICP queries in a timely manner and also sporadically refusing
>TCP connections on the proxy port.
>
>--
>Henrik Nordstrom
>Squid hacker
>
>
Received on Fri Oct 22 1999 - 01:41:02 MDT