Re: [squid-users] COSS still under developement?

From: Tek Bahadur Limbu <teklimbu@dont-contact.us>
Date: Thu, 4 Oct 2007 11:17:43 +0545

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Hi Leongmzlist,

On Wed, 03 Oct 2007 14:57:17 -0700
leongmzlist <leongmzlist@gmail.com> wrote:

> Hi,
>
> The list has been very quiet w/ regards to the coss storage.
>
> 1. is it still being developed/maintained?
> 2. is it production ready?

I am using COSS for some of my FreeBSD Squid proxy servers. The reason why I chose the COSS storage scheme was because DISKD kept on crashing repeatedly under high load (150 - 200 req/sec).

I have to say that it's quite stable. Squid hasn't crashed in last 35 days since I deployed COSS in it.

It seems quite fast too keeping in mind that we have a satellite link.

Below are some of stats after deploying COSS in one of my proxies:

- ----------------------------------------------------------------

root# squidclient mgr:info

HTTP/1.0 200 OK
Server: squid
Date: Thu, 04 Oct 2007 05:00:43 GMT
Content-Type: text/plain
Expires: Thu, 04 Oct 2007 05:00:43 GMT
Last-Modified: Thu, 04 Oct 2007 05:00:43 GMT
X-Cache: MISS from cache2-server
X-Cache-Lookup: MISS from cache2-server:3128
Via: 1.0 cache2-server:3128 (squid)
Proxy-Connection: close

Squid Object Cache: Version 2.6.STABLE14
Start Time: Fri, 31 Aug 2007 10:24:34 GMT
Current Time: Thu, 04 Oct 2007 05:00:43 GMT
Connection information for squid:
        Number of clients accessing cache: 4168
        Number of HTTP requests received: 173803166
        Number of ICP messages received: 1090974817
        Number of ICP messages sent: 1095742664
        Number of queued ICP replies: 9
        Request failure ratio: 0.00
        Average HTTP requests per minute since start: 3573.5
        Average ICP messages per minute since start: -43347.4
        Select loop called: -1251324801 times, -2.332 ms avg
Cache information for squid:
        Request Hit Ratios: 5min: 50.6%, 60min: 50.2%
        Byte Hit Ratios: 5min: 27.6%, 60min: 29.1%
        Request Memory Hit Ratios: 5min: 0.0%, 60min: 0.0%
        Request Disk Hit Ratios: 5min: 74.3%, 60min: 74.7%
        Storage Swap size: 11074963 KB
        Storage Mem size: 124912 KB
        Mean Object Size: 9.27 KB
        Requests given to unlinkd: 0
Median Service Times (seconds) 5 min 60 min:
        HTTP Requests (All): 0.04047 0.10281
        Cache Misses: 1.46131 1.46131
        Cache Hits: 0.00091 0.00091
        Near Hits: 1.17732 1.17732
        Not-Modified Replies: 0.00091 0.00091
        DNS Lookups: 0.06083 0.06364
        ICP Queries: 0.00102 0.00108
Resource usage for squid:
        UP Time: 2918169.381 seconds
        CPU Time: 325875.751 seconds
        CPU Usage: 11.17%
        CPU Usage, 5 minute avg: 23.26%
        CPU Usage, 60 minute avg: 25.13%
        Process Data Segment Size via sbrk(): 789776 KB
        Maximum Resident Size: 637880 KB
        Page faults with physical i/o: 871
Memory accounted for:
        Total accounted: 439005 KB
        memPoolAlloc calls: 2618284649
        memPoolFree calls: 2613332187
File descriptor usage for squid:
        Maximum number of file descriptors: 8192
        Largest file desc currently in use: 1969
        Number of file desc currently in use: 1583
        Files queued for open: 0
        Available number of file descriptors: 6609
        Reserved number of file descriptors: 100
        Store Disk files open: 0
        IO loop method: kqueue
Internal Data Structures:
        1195109 StoreEntries
           741 StoreEntries with MemObjects
            40 Hot Object Cache Items
        1194406 on-disk objects

The reason there is high number of page faults is due to 2 long outages we had last month!.

I am also a little puzzled by the negative values of the Select loop called!

- ---------------------------------------------------

root# squidclient mgr:storedir

HTTP/1.0 200 OK
Server: squid
Date: Thu, 04 Oct 2007 05:07:02 GMT
Content-Type: text/plain
Expires: Thu, 04 Oct 2007 05:07:02 GMT
Last-Modified: Thu, 04 Oct 2007 05:07:02 GMT
X-Cache: MISS from cache2-server
X-Cache-Lookup: MISS from cache2-server:3128
Via: 1.0 cache2-server:3128 (squid)
Proxy-Connection: close

Store Directory Statistics:
Store Entries : 1195347
Maximum Swap Size : 12582912 KB
Current Store Swap Size: 11073264 KB
Current Capacity : 88% used, 12% free

Store Directory #0 (coss): /cache1/squid/coss
FS Block Size 1024 Bytes

Maximum Size: 6291456 KB
Current Size: 5536468 KB
Percent Used: 88.00%
Current load metric: 0 / 1000
Number of object collisions: 2
Flags:
Pending Relocations: 0
Stripe: 1180, lockcount: 0, numobjects 40, flags: NOTFULL,NOTWRITING,NOTWRITTEN,DISK
Stripe: 1, lockcount: 4, numobjects 204, flags: NOTFULL,NOTWRITING,NOTWRITTEN,MEMONLY
Stripe: 1178, lockcount: 1, numobjects 124, flags: FULL,NOTWRITING,NOTWRITTEN,DISK
Stripe: 11, lockcount: 1, numobjects 190, flags: FULL,NOTWRITING,NOTWRITTEN,MEMONLY
Stripe: 8, lockcount: 1, numobjects 189, flags: FULL,NOTWRITING,NOTWRITTEN,MEMONLY
Stripe: 4, lockcount: 1, numobjects 175, flags: FULL,NOTWRITING,NOTWRITTEN,MEMONLY
Stripe: 6, lockcount: 1, numobjects 175, flags: FULL,NOTWRITING,NOTWRITTEN,MEMONLY
Stripe: 1161, lockcount: 1, numobjects 121, flags: FULL,NOTWRITING,NOTWRITTEN,DISK
Stripe: 2, lockcount: 1, numobjects 145, flags: FULL,NOTWRITING,NOTWRITTEN,MEMONLY
Stripe: 5, lockcount: 1, numobjects 182, flags: FULL,NOTWRITING,NOTWRITTEN,MEMONLY
Stripe: 0, lockcount: 1, numobjects 135, flags: FULL,NOTWRITING,NOTWRITTEN,MEMONLY
Pending Relocations:
Accepted object sizes: 0 - 131072 bytes

Store Directory #1 (coss): /cache2/squid/coss
FS Block Size 1024 Bytes

Maximum Size: 6291456 KB
Current Size: 5536796 KB
Percent Used: 88.01%
Current load metric: 0 / 1000
Number of object collisions: 3
Flags:
Pending Relocations: 1
Stripe: 13, lockcount: 5, numobjects 142, flags: NOTFULL,NOTWRITING,NOTWRITTEN,MEMONLY
Stripe: 12, lockcount: 2, numobjects 132, flags: FULL,NOTWRITING,NOTWRITTEN,MEMONLY
Stripe: 6, lockcount: 1, numobjects 134, flags: FULL,NOTWRITING,NOTWRITTEN,MEMONLY
Stripe: 4, lockcount: 3, numobjects 122, flags: FULL,NOTWRITING,NOTWRITTEN,MEMONLY
Stripe: 4261, lockcount: 0, numobjects 117, flags: NOTFULL,NOTWRITING,NOTWRITTEN,DISK
Stripe: 11, lockcount: 1, numobjects 149, flags: FULL,NOTWRITING,NOTWRITTEN,MEMONLY
Stripe: 10, lockcount: 1, numobjects 154, flags: FULL,NOTWRITING,NOTWRITTEN,MEMONLY
Stripe: 7, lockcount: 1, numobjects 156, flags: FULL,NOTWRITING,NOTWRITTEN,MEMONLY
Stripe: 5, lockcount: 1, numobjects 138, flags: FULL,NOTWRITING,NOTWRITTEN,MEMONLY
Stripe: 0, lockcount: 2, numobjects 116, flags: FULL,NOTWRITING,NOTWRITTEN,MEMONLY
Stripe: 2, lockcount: 1, numobjects 113, flags: FULL,NOTWRITING,NOTWRITTEN,MEMONLY
Stripe: 8, lockcount: 1, numobjects 120, flags: FULL,NOTWRITING,NOTWRITTEN,MEMONLY
Stripe: 1, lockcount: 1, numobjects 159, flags: FULL,NOTWRITING,NOTWRITTEN,MEMONLY
Stripe: 9, lockcount: 1, numobjects 129, flags: FULL,NOTWRITING,NOTWRITTEN,MEMONLY
Pending Relocations:
  Stripe: 3968 Number: 1
Accepted object sizes: 0 - 131072 bytes

- -----------------------------------------------------------------------

squidclient mgr:5min | grep client

client_http.requests = 142.789173/sec
client_http.hits = 72.886244/sec
client_http.errors = 0.000000/sec
client_http.kbytes_in = 104.722727/sec
client_http.kbytes_out = 1241.499474/sec
client_http.all_median_svc_time = 0.036222 seconds
client_http.miss_median_svc_time = 1.542425 seconds
client_http.nm_median_svc_time = 0.000911 seconds
client_http.nh_median_svc_time = 1.177316 seconds
client_http.hit_median_svc_time = 0.000911 seconds

- -----------------------------------------------

I need further suggestions to refine and fine-tune COSS for my systems.
I appreciate suggestions from anybody regarding my stats above.

I definitely hope that COSS will become better in the future.

Thanking you...

>
> mike
>
>

- --

With best regards and good wishes,

Yours sincerely,

Tek Bahadur Limbu

System Administrator

(TAG/TDG Group)
Jwl Systems Department

Worldlink Communications Pvt. Ltd.

Jawalakhel, Nepal
http://wlink.com.np/

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.7 (FreeBSD)

iD8DBQFHBHr4fpE0pz+xqQQRAoGIAKC/6qqyGY+ELbHiGMBoOb9djhKCdgCfUWQx
yG9SJMQYEo9g1QaCGgI919g=
=SxY9
-----END PGP SIGNATURE-----
Received on Wed Oct 03 2007 - 23:33:14 MDT

This archive was generated by hypermail pre-2.1.9 : Thu Nov 01 2007 - 13:00:01 MDT