[squid-users] Re: cache_dir amendment

From: Irwin Soo <[email protected]>
Date: Wed, 28 Nov 2001 13:57:10 +0800

I did some changes by changing

1) File Descriptors
/* Number of descriptors that can fit in an `fd_set'. */
#define __FD_SETSIZE 8192

2) ulimit prior to squid startup
ulimit -HSn 16384

restart the cache box cleanly, but cache.log shows

2001/11/28 13:43:57| Starting Squid Cache version 2.4.STABLE2 for i686-pc-linux-gnu...
2001/11/28 13:43:57| Process ID 729
2001/11/28 13:43:57| With 1024 file descriptors available
2001/11/28 13:43:57| DNS Socket created on FD 4
2001/11/28 13:43:57| helperOpenServers: Starting 10 'squirm' processes
2001/11/28 13:43:57| Unlinkd pipe opened on FD 19
2001/11/28 13:43:57| Swap maxSize 50331648 KB, estimated 3871665 objects
2001/11/28 13:43:57| Target number of buckets: 193583
2001/11/28 13:43:57| Using 262144 Store buckets
2001/11/28 13:43:57| Max Mem size: 65536 KB
2001/11/28 13:43:57| Max Swap size: 50331648 KB
2001/11/28 13:43:57| Local cache digest enabled; rebuild/rewrite every 3600/3600 sec
2001/11/28 13:43:57| Rebuilding storage in /directory1 (DIRTY)
2001/11/28 13:43:57| Rebuilding storage in /directory2 (DIRTY)
2001/11/28 13:43:57| Rebuilding storage in /directory3 (DIRTY)
2001/11/28 13:43:57| Rebuilding storage in /directory4 (DIRTY)
2001/11/28 13:43:57| Rebuilding storage in /directory5 (DIRTY)
2001/11/28 13:43:57| Rebuilding storage in /directory6 (DIRTY)
2001/11/28 13:43:57| Using Least Load store dir selection
2001/11/28 13:43:57| Set Current Directory to /directory1
2001/11/28 13:43:57| Loaded Icons.
2001/11/28 13:43:57| Accepting HTTP connections at 0.0.0.0, port 8080, FD 36.
2001/11/28 13:43:57| Accepting ICP messages at 0.0.0.0, port 3130, FD 38.
2001/11/28 13:43:57| Accepting SNMP messages on port 3401, FD 39.
2001/11/28 13:43:57| WCCP Disabled.
2001/11/28 13:43:57| WARNING: Peer looks like this host
2001/11/28 13:43:57| Ready to serve requests.
2001/11/28 13:43:57| Done reading /directory6 swaplog (2397 entries)
2001/11/28 13:43:57| Done reading /directory2 swaplog (2509 entries)
2001/11/28 13:43:57| Done reading /directory3 swaplog (2732 entries)
2001/11/28 13:43:57| Done reading /directory4 swaplog (2753 entries)
2001/11/28 13:43:57| Done reading /directory1 swaplog (2829 entries)
2001/11/28 13:43:57| Done reading /directory5 swaplog (3264 entries)
2001/11/28 13:43:57| Finished rebuilding storage from disk.
2001/11/28 13:43:57| 15491 Entries scanned
2001/11/28 13:43:57| 0 Invalid entries.
2001/11/28 13:43:57| 0 With invalid flags.
2001/11/28 13:43:57| 14526 Objects loaded.
2001/11/28 13:43:57| 0 Objects expired.
2001/11/28 13:43:57| 931 Objects cancelled.
2001/11/28 13:43:57| 64 Duplicate URLs purged.
2001/11/28 13:43:57| 4 Swapfile clashes avoided.
2001/11/28 13:43:57| Took 0.0 seconds (14526.0 objects/sec).
2001/11/28 13:43:57| Beginning Validation Procedure
2001/11/28 13:43:57| Completed Validation Procedure
2001/11/28 13:43:57| Validated 14492 Entries
2001/11/28 13:43:57| store_swap_size = 200396k
2001/11/28 13:44:04| storeLateRelease: released 0 objects

not sure what really went wrong with respect to :

1)2001/11/28 13:43:57| With 1024 file descriptors available
   - which I thought that I would see 8192?

2) 2001/11/28 13:43:57| Rebuilding storage in /directory1 (DIRTY)
2001/11/28 13:43:57| Rebuilding storage in /directory2 (DIRTY)
2001/11/28 13:43:57| Rebuilding storage in /directory3 (DIRTY)
2001/11/28 13:43:57| Rebuilding storage in /directory4 (DIRTY)
2001/11/28 13:43:57| Rebuilding storage in /directory5 (DIRTY)
2001/11/28 13:43:57| Rebuilding storage in /directory6 (DIRTY)

- which I restart squid cleanly?

pse help! tks.

Rgds,

Irwin

  ----- Original Message -----
  From: Irwin Soo
  To: squid-users@squid-cache.org
  Sent: Tuesday, November 27, 2001 10:30 PM
  Subject: cache_dir amendment

  Hi all,

  I'm a newbie here...

  I 'm running squid2.4stable2 on RH7.1, here are some of my config before posing my questions here:-

  1) cache partition
  /dev/sdb1 53G 542M 49G 2% /directory1
  /dev/sdb5 53G 542M 49G 2% /directory2
  /dev/sdb6 53G 544M 49G 2% /directory3
  /dev/sdb7 53G 543M 49G 2% /directory4
  /dev/sdb8 53G 542M 49G 2% /directory5
  /dev/sdb9 53G 542M 49G 2% /directory6

  2) File Descriptors
  /* Number of descriptors that can fit in an `fd_set'. */
  #define __FD_SETSIZE 16384

  3) ulimit prior to squid startup
  ulimit -HSn 32768

  4) max limit in /etc/sysctl.conf
  # Increase file_max(/proc/sys/fs/file-max) size automatically
  fs.file-max = 32768

  Q: any serious misconfig from the above?

  In my squid.conf

  #Default:
  # cache_dir ufs /usr/local/squid/cache 100 16 256
  # cache_dir diskd /usr/local/squid/cache 8192 128 1024 Q1=72 Q2=96
   cache_dir diskd /directory1 8192 128 1024 Q1=72 Q2=96
   cache_dir diskd /directory2 8192 128 1024 Q1=72 Q2=96
   cache_dir diskd /directory3 8192 128 1024 Q1=72 Q2=96
   cache_dir diskd /directory4 8192 128 1024 Q1=72 Q2=96
   cache_dir diskd /directory5 8192 128 1024 Q1=72 Q2=96
   cache_dir diskd /directory6 8192 128 1024 Q1=72 Q2=96

  with each physical cache partition to be 49GB, seemed like I only allocated 8GB out of each,
  with directory and sub-directory to be 128 and 1024 respectively.

  If I want to rectify my cache directory, is reinstallation the only resort? or I can perform steps
  e.g. changes to cache_dir in squid.conf, recompile, restart squid as a alternative to reinstallation?
         

  # Q1 specifies the number of unacknowledged I/O requests when Squid
  # stops opening new files. If this many messages are in the queues,
  # Squid won't open new files. Default is 64
  #
  # Q2 specifies the number of unacknowledged messages when Squid
  # starts blocking. If this many messages are in the queues,
  # Squid blocks until it recevies some replies. Default is 72

  What are the considerations besides using the default values to directories, subdirectories and Q1/Q2 settings
  taking into consideration read/write optimisation? I'm struggling to understand the description of Q!/Q2 in simpler terms.
  appreciate your help here.

  p.s. my h/w spec.(s) Intel PIII 1GB, 2GB RAM, 5x73GB cache data allocation(exclusive of OS, RAID5)

  Thank you.

  Rgds,

  Irwin
Received on Tue Nov 27 2001 - 22:57:14 MST

This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 17:04:34 MST