Re: High load squid server

From: Duane Wessels <wessels@dont-contact.us>
Date: Sun, 25 Jun 2000 22:28:56 -0600

On Mon, 26 Jun 2000, Wisnu Wijayanta wrote:

> Hi all,
>
> I set up this tiny... :) server to accomodate our corporation internet
> access.
>
> Sever - Now:
> Caldera e-Server 2.3,
> Pentium Pro 200 MMX,
> with 160 Mb RAM,
> 4 * 4 Gb SCSI-2,
> Raid 0 (I use this option just because the smartarray is already installed
> and I only have one dual channel scsi controller)
>
> Squid:
> I use dlmalloc, async-io, LFUDA, quite large acl and regex configuration, 10
> Gb disk for cache directory.
>
> Recently, I notice that CPU usage is considerably high. And next month, I
> want to replace this "old but nice" box, with single Xeon-866, 1 Gb RAM, two
> dual channel W-U3 SCSI-controllers which is 4*18 Gb W-U3 SCSI 10K rpm disk
> will be installed.
>
> Here, I want to ask some question before I move to the new-box, first why
> total memory accounted from the cachemgr isn't the same with the result by
> "top"?

http://www.squid-cache.org/Doc/FAQ/FAQ-8.html#ss8.6

> Second, what is the meaning of Storage Replacement Treshold?

http://www.squid-cache.org/Doc/FAQ/FAQ-12.html#ss12.12

> Third,
> are there any disk partitioning consideration regarding with squid
> performance?

I recommend one filesystem per disk drive. Never RAID.

> Last, are the CPU load normal, do i need to think about dual
> CPU options?

Your CPU usage seems abnormally high for your request rate (7/sec).

I wonder if you applied this patch?
http://www.squid-cache.org/Versions/v2/2.3/bugs/#squid-2.3.stable3-storeExpiredReferenceAge

> Any help would be very appreciated.
>
> Wisnu.
>
> ps. this is my squid report.
>
> Squid Object Cache: Version 2.3.STABLE3
>
> Start Time: Fri, 16 Jun 2000 08:01:56 GMT
> Current Time: Mon, 26 Jun 2000 02:24:40 GMT
>
> Connection information for squid:
> Number of clients accessing cache: 1717
> Number of HTTP requests received: 5910210
> Number of ICP messages received: 289076
> Number of ICP messages sent: 289076
> Number of queued ICP replies: 0
> Request failure ratio: 0.00%
> HTTP requests per minute: 420.3
> ICP messages per minute: 41.1
> Select loop called: 58023024 times, 14.542 ms avg
> Cache information for squid:
> Request Hit Ratios: 5min: 48.9%, 60min: 39.6%
> Byte Hit Ratios: 5min: 26.1%, 60min: 28.8%
> Storage Swap size: 7623294 KB
> Storage Mem size: 32744 KB
> Storage Replacement Threshold: 0.000000
> Mean Object Size: 11.09 KB
> Requests given to unlinkd: 0
> Median Service Times (seconds) 5 min 60 min:
> HTTP Requests (All): 0.17711 0.68577
> Cache Misses: 2.13280 1.81376
> Cache Hits: 0.12106 0.18699
> Near Hits: 2.25116 1.91442
> Not-Modified Replies: 0.05633 0.10857
> DNS Lookups: 1.00974 0.64769
> ICP Queries: 0.00000 0.00000
> Resource usage for squid:
> UP Time: 843764.093 seconds
> CPU Time: 204409.390 seconds
> CPU Usage: 24.23%
> CPU Usage, 5 minute avg: 78.55%
> CPU Usage, 60 minute avg: 75.72%
> Maximum Resident Size: 0 KB
> Page faults with physical i/o: 2478173
> Memory usage for squid via mallinfo():
> Total space in arena: 129117 KB
> Ordinary blocks: 119900 KB 149283 blks
> Small blocks: 0 KB 0 blks
> Holding blocks: 3180 KB 2 blks
> Free Small blocks: 0 KB
> Free Ordinary blocks: 9216 KB
> Total in use: 123080 KB 95%
> Total free: 9216 KB 7%
> Memory accounted for:
> Total accounted: 80789 KB
> File descriptor usage for squid:
> Maximum number of file descriptors: 1024
> Largest file desc currently in use: 395
> Number of file desc currently in use: 342
> Files queued for open: 0
> Available number of file descriptors: 682
> Reserved number of file descriptors: 100
> Store Disk files open: 9
> Internal Data Structures:
> 694517 StoreEntries
> 7679 StoreEntries with MemObjects
> 7581 Hot Object Cache Items
> 687253 Filemap bits set
> 687252 on-disk objects
>
Received on Sun Jun 25 2000 - 22:33:13 MDT

This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:54:10 MST