First, let's look at the squid.conf options available for hierarchy configuration. We will then work through the most common hierarchy structures, so that you can see the way that the options are used.
You use the cache_peer option to configure the peers that Squid will communicate with. Other options are then used to select which peer to pass a request to.
When communicating with a peer, Squid needs some basic information about how to talk to the machine; the hostname, what ports to send queries to, and so forth. The cache_peer config line does this. Let's look at an example line:
The cache_peer option is split into five fields. The first field (cache.domain.example) is the hostname or IP of the cache that is to be queried. The second field indicates the type of relationship, and must be set to either parent or sibling or multicast. The third field sets the HTTP port of the destination server, while the fourth sets the ICP (UDP) query port. The fifth field can contain more than zero or more keywords, although we only use one in the example above; the keyword default sets that the cache will be used as the default path to the outside world. If you compiled Squid to support HTCP, your cache will automatically attempt to connect to TCP port 4827 (there is currently no option to change this port value). Cache digests are transferred via the HTTP port specified on the cache_peer line.
Here is a summary of the available cache_peer options:
proxy-only. Data retrieved from this remote cache will not be stored locally, but retrieved again on any subsequent request. By default Squid will store objects it retrieves from other caches: by having the object available locally it can return the object fast if it's ever requested again. While this is good for latency, it can be a waste of bandwidth, especially if the other cache is on the same piece of ethernet. In the examples section of this chapter, we use this option when load-balancing between two cache servers.
weight. If more than one cache server has an object (based on the result of an ICP query), Squid decides which cache to get the data from the cache that responded fastest. If you want to prefer one cache over another, you can add a weight value to the preferred cache's config line. Larger values are preferred. Squid times how long each ICP request takes (in milliseconds), and divides the time by the weight value, using the cache with the smallest result. Your weight value should thus not be an unreasonable value.
ttl. This tag is covered in the multicast section, later in this chapter.
no-query. Squid will send ICP requests to all configured caches. The response time is measured, and used to decide which parent to send the HTTP request to. There is another function of these requests: if there is no response to a request, the cache is marked down. If you are communicating with a cache that does not support ICP, you must use the no-query option: if you don't, Squid will consider that cache down, and attempt to go directly to the destination server. (If you want, you can set the ICP port on the config line to point to the echo port, port 7. Squid will then use this port to check if the machine is available. Note that you will have to configure inetd.conf to support the UDP echo port.) This option is normally used in conjunction with the default option.
default. This sets the host to be the proxy of last resort. If no other cache matches a rule (due to acl or domain filtering), this cache is used. If you have only one way of reaching the outside world, and it doesn't support ICP, you can use the default and no-query options to ensure that all queries are passed through it. If this cache is then down, the client will see an error message (without these options, Squid would attempt to route around the problem.)
round-robin. This option must be used on more than one cache_peer line to be useful. Connections to caches configured with this options are spread evenly (round-robined) among the caches. This can be used by client caches to communicate with a group of loaded parents, so that load is spread evenly. If you have multiple Internet connections, with a parent cache on each side, you can use this option to do some basic load-balancing of the connections.
multicast-responder. This option is covered in the multicast section later in this chapter.
closest-only.
no-netdb-exchange. If your cache was configured to keep ICMP (ping) timing information with the --enable-icmp configure option, your cache will attempt to retrieve the remote machine's ICMP timing information from any peers. If you don't want this to happen (or the remote cache doesn't support it), you can use the no-netdb-exchange option to stop Squid from requesting this information from the cache.
no-delay. Hits from other caches will normally be included into a client's delay-pool information. If you have two caches load-balancing, you don't want the hits from the other cache to be limited. You may also want hits from caches in a nearby hierarchy to come down at full speed, not to be limited as if they were misses. Use the no-delay option to ensure that requests come down at their full speed.
login. Caches can be configured to use usernames and passwords on accesses. To authenticate with a parent cache, you can enter a username and password using this tag. Note that the HTTP protocol makes authenticating to multiple cache servers impossible: you cannot chain together a string of proxies, each one requiring authentication. You should only use this option if this is a personal proxy.