Re: AW: [squid-users] Howto Clear Cache Periodicaly

On Thu, Apr 13, 2006, nonama wrote:
>
> Thank you. I adapt some of the config from default. I
> have commented it out, surfing seems a bit faster.
> However still need some help on the cache.

I'm coming into this a bit late so I apologise if I ask
for anything you've already covered.

* What kind of machine are you running it on? (RAM/CPU)
* What kind of disks?
* Is the machine paging/swapping? ("vmstat 1" on Linux servers
  will give a general indication how often its paging memory
  to/from disk.)
* How many requests a minute is going through the machine?
  The cache manager cgi (cachemgr.cgi) or the 'squidclient' program
  can give you an indication. Try this on the squid server:

<host>:/usr/local/src/squid-monotone/net.cacheboy.squid.ucs-local/src# ./squidclient mgr:info | head -30
HTTP/1.0 200 OK
Server: squid/2.5.STABLE13-CVS-CACHEBOY
Mime-Version: 1.0
Date: Thu, 13 Apr 2006 07:52:14 GMT
Content-Type: text/plain
Expires: Thu, 13 Apr 2006 07:52:14 GMT
Last-Modified: Thu, 13 Apr 2006 07:52:14 GMT
X-Cache: MISS from <host>
Proxy-Connection: close

Squid Object Cache: Version 2.5.STABLE13-CVS-CACHEBOY
Start Time: Thu, 13 Apr 2006 06:17:32 GMT
Current Time: Thu, 13 Apr 2006 07:52:14 GMT
Connection information for squid:
        Number of clients accessing cache: 1629
        Number of HTTP requests received: 563029
        Number of ICP messages received: 435155
        Number of ICP messages sent: 435655
        Number of queued ICP replies: 0
        Request failure ratio: 0.00
        Average HTTP requests per minute since start: 5946.3
        Average ICP messages per minute since start: 9196.8
        Select loop called: 64726368 times, 0.088 ms avg
Cache information for squid:
        Request Hit Ratios: 5min: 22.7%, 60min: 20.5%
        Byte Hit Ratios: 5min: 14.3%, 60min: 13.1%
        Request Memory Hit Ratios: 5min: 3.9%, 60min: 6.4%
        Request Disk Hit Ratios: 5min: 37.4%, 60min: 39.1%
        Storage Swap size: 47923640 KB
        Storage Mem size: 98268 KB

What you're after is the "Select loop called" average time (should be under
20 milliseconds) and the average HTTP requests per minute.

It might be a few things;

* You might find you're running out of RAM and its paging to disk;
* You might be putting a a lot of requests through squid and need
  to move to aufs or diskd
* Other stuff

Adrian
Received on Thu Apr 13 2006 - 01:55:46 MDT