Squid_ The Definitive Guide - Duane Wessels [147]
The amount of traffic received from neighbors in ICP queries.
icp.r_kbytes_recv
The amount of traffic received from neighbors in ICP replies.
icp.times_used
The number of times ICP resulted in the selection of a neighbor as the next-hop for a cache miss.
cd.times_used
The number of times Cache Digests resulted in the selection of a neighbor as the next-hop for a cache miss.
cd.msgs_sent
The number of Cache Digest messages sent to neighbors.
cd.msgs_recv
The number of Cache Digest messages received from neighbors.
cd.memory
The amount of memory (in kilobytes) used by enabling the Cache Digests' feature.
cd.local_memory
The amount of memory (in kilobytes) used to store Squid's own Cache Digest.
cd.kbytes_sent
The amount of traffic sent to neighbors in Cache Digest messages.
cd.kbytes_recv
The amount of traffic received from neighbors in Cache Digest messages.
unlink.requests
The number of unlink requests given to the (optional) unlinkd process.
page_faults
The number of (major) page faults as reported by getrusage( ).
select_loops
The number of times Squid called select( ) or poll( ) in the main I/O loop.
cpu_time
The amount of CPU time (in seconds) accumulated, as reported by getrusage( ).
wall_time
The amount of human time (in seconds) elapsed since Squid was started.
swap.outs
The number of objects (swap files) saved to disk.
swap.ins
The number of objects (swap files) read from disk.
swap.files_cleaned
The number of orphaned cache files removed by the periodic cleanup procedure.
aborted_requests
The number of server-side HTTP requests aborted due to client-side aborts.
peer_select: Peer Selection Algorithms
This page contains a lot of low-level detail about cache digests that I won't discuss. Most of the numbers are meaningful only to the developers that originally wrote the Cache Digest implementation.
However, at the end of this page is a little table that compares Algorithm usage:
Algorithm usage:
Cache Digest: 27 ( 24%)
Icp: 84 ( 76%)
Total: 111 (100%)
In this example, you can see that Squid sent 111 requests to one of its neighbors: 27 are due to Cache Digests and 84 are due to ICP. In this context, ICP also includes HTCP.
digest_stats: Cache Digest and ICP Blob
This page is actually just a concatenation of the following other cache manager pages:
Traffic and Resource Counters
5 Minute Average of Counters
Full Histogram Counts
Peer Selection Algorithms
Store Digest
Its only purpose is to enable developers to take a snapshot of a number of statistics with a single request.
5min: 5 Minute Average of Counters
This page shows a five-minute average of the data in the Traffic and Resource Counters page. In addition to the counters mentioned in Section 14.2.1.30, this page also contains the following values:
client_http.all_median_svc_time
The median service (response) time for all client requests from the last five minutes.
client_http.miss_median_svc_time
The median service time for cache misses from the last five minutes.
client_http.nm_median_svc_time
The five-minute median service time for requests logged as TCP_IMS_HIT. See "Not-Modified Replies" in Section 14.2.1.24.
client_http.nh_median_svc_time
The five-minute median service time for Near Hits (TCP_REFRESH_HIT requests).
client_http.hit_median_svc_time
The five-minute median service time for unvalidated cache hits.
icp.query_median_svc_time
The five-minute median service time for ICP queries sent by Squid (how long it takes for the neighbors to reply to our queries).
icp.reply_median_svc_time
The five-minute median service time for ICP queries received by Squid (how long it takes Squid to reply to its neighbor's queries). ICP processing normally occurs faster than the process clock is updated, so this value is always zero.
dns.median_svc_time
The five-minute median service time for DNS queries.
select_fds
The mean rate at which the main I/O loop scans file descriptors with select( ) or poll( ). Note: a low number doesn't necessarily