[SQU] fd_table[97].rwstate + Slow Cache.

From: Pat Newby <P.Newby@dont-contact.us>
Date: Tue, 12 Sep 2000 17:44:47 +0100 (BST)

Squid 2.2.STABLE4

Two questions:

(1) Can someone tell me what these messages in cache.log mean, please?
      comm_write: fd_table[97].rwstate != NULL

    Every so often we get several of these a second.

(2)

I'm trying to track down a problem with our cache which, since Friday,
has been producing core dumps about every other day, and going on go-slows
every few hours - eg 20 seconds instead of 1 second to get a page from a
parent, and 3 seconds instead of milliseconds to get a page from local
disk. This is cured by:
  squid -k shutdown
and letting RunCache restart it.

The usage is about a third of usual, because the students are away.

After several months of having a cache on a single 4gb disk, I added a
second cache disk of 18gb, with 16.5 gb allocated to the cache. I changed
nothing in squid.conf except to add the second cache_dir line:
   cache_dir /wcache/squid/cache 3800 16 256
   cache_dir /wcache1/squid/cache 16500 48 256
This was on 22nd August, and it ran very well until Friday.

Is it coincidence that on Friday the second cache disk reached 92% full,
with cache space used, out of cache space allocated, being 91% ? I now
have both caches at 91.07% used, ie equilibrium.

Before Friday, the cache was writing every new page to the new disk (I
think). From Friday, new pages have been written to both disks, to many
directories, overwriting old files. Is this significant ?

Can anyone suggest what to look at ? There's nothing in cache.log except
      comm_write: fd_table[97].rwstate != NULL
which we have had before, but perhaps not so many times.

Pat.
Pat Newby, IT and Computing Services (ITCS),
           University of East Anglia, Norwich, NR4 7TJ, UK.

--
To unsubscribe, see http://www.squid-cache.org/mailing-lists.html
Received on Tue Sep 12 2000 - 10:48:01 MDT

This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:55:16 MST