Making main squid cache fail over better

From: Clifton Royston <cliftonr@dont-contact.us>
Date: Wed, 1 Dec 1999 15:34:50 -1000

  OK, I've finally got approval and budget to press forward on doing a
large-scale web cache for our customers, with around 54GB disk for
cache, 512MB of RAM in the server, etc. We want to make use of the
cache voluntary, and give people incentive to use it. We won't use
transparent proxy, at least for a while. So, one of the big factors in
whether people will use it is whether it's easy to set up and highly
reliable. I can deal with the first with all the help people have
given about proxy.pac scripts.

  However... if we do get thousands of users pointing their browsers to
it, we don't want to get thousands of phone calls if one of the disks
crashes. There have been lots of comments against using RAID 5 for
performance reasons, and using RAID 1 would double the disk cost.

  What have other Squid users been able to do to ensure fast and
preferably automatic recovery from a disk failure or disk crash, short
of using RAID? Has anybody written a script which would (e.g.) detect
that a disk has been taken offline (or should be, due to errors),
comment out the corresponding cache directory from the Squid config,
shut it down, and restart it?

  The other alternative I'm considering is to have two front-end
servers which all of the clients would point their browsers at, using
an L4 switch to distribute load between them and handle fail-over.
These would have essentially no disk and put their minimal cache in
RAM, both using the disk-based Squid as a parent, with cache digests,
and failing over to go direct. This would probably end up cheaper and
more robust than one server with RAID-0 on the disk, and would give us
the invisible failover of the better transparent proxy schemes.
However, it does mean two more servers to worry about maintaining.

  Any comments, suggestions, ideas, bricks? Ideas on getting people to
use it - would an ad-zapper setting on the proxy be a good incentive?
  -- Clifton

-- 
 Clifton Royston  --  LavaNet Systems Architect --  cliftonr@lava.net
        "An absolute monarch would be absolutely wise and good.  
           But no man is strong enough to have no interest.  
             Therefore the best king would be Pure Chance.  
              It is Pure Chance that rules the Universe; 
          therefore, and only therefore, life is good." - AC
Received on Wed Dec 01 1999 - 18:40:52 MST

This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:49:41 MST