Re: [SQU] The old cache over limit problem

On Thu, 16 Nov 2000, Duane Wessels wrote:

>
>
> On Thu, 16 Nov 2000, Ethy H. Brito wrote:
>
> >
> > Dear Squidere
> >
> > The thing is getting worse here! Almost no light at the end of the tunnel!
> >
> > servidor:/cache# df
> > Filesystem 1k-blocks Used Available Use% Mounted on
> > /dev/hda8 1011928 955084 5440 99% /cache
> >
> > Even after I have changed my replacement_policy from LFUDA to LRU. Squid
> > refuses to delete files! (obviouly I restarted it!)
>
> did you show us your 'client mgr:storedir' output?

HTTP/1.0 200 OK
Server: Squid/2.3.STABLE4
Mime-Version: 1.0
Date: Thu, 16 Nov 2000 17:11:38 GMT
Content-Type: text/plain
Expires: Thu, 16 Nov 2000 17:11:38 GMT
Last-Modified: Thu, 16 Nov 2000 17:11:38 GMT
X-Cache: MISS from xxx
Proxy-Connection: close

Store Directory Statistics:
Store Entries : 97641
Maximum Swap Size : 819200 KB
Current Store Swap Size: 703584 KB
Current Capacity : 86% used, 14% free

Store Directory #0: /cache
First level subdirectories: 16
Second level subdirectories: 256
Maximum Size: 819200 KB
Current Size: 703584 KB
Percent Used: 85.89%
Filemap bits in use: 97633 of 131072 (74%)
Filesystem Space in use: 908728/1011928 KB (90%)
Filesystem Inodes in use: 101757/128768 (79%)
Flags:

I don't know if will be of any help but:

2000/11/16 14:59:11| Done scanning /cache swaplog (0 entries)
2000/11/16 14:59:11| Finished rebuilding storage from disk.
2000/11/16 14:59:11| 97606 Entries scanned
2000/11/16 14:59:11| 0 Invalid entries.
2000/11/16 14:59:11| 0 With invalid flags.
2000/11/16 14:59:11| 97606 Objects loaded.
2000/11/16 14:59:11| 0 Objects expired.
2000/11/16 14:59:11| 0 Objects cancelled.
2000/11/16 14:59:11| 0 Duplicate URLs purged.
2000/11/16 14:59:11| 0 Swapfile clashes avoided.
2000/11/16 14:59:11| Took 693.3 seconds ( 140.8 objects/sec).
2000/11/16 14:59:11| Beginning Validation Procedure
2000/11/16 14:59:12| Completed Validation Procedure
2000/11/16 14:59:12| Validated 97606 Entries
2000/11/16 14:59:12| store_swap_size = 703452k
2000/11/16 14:59:13| storeLateRelease: released 1 objects

store_swap_size is reported to be 703452k but df shows other thing (see below)

>
> does cache.log complain about being over the limit?

Nop. Squid dies after running out of space on file system.

And Squid knows file system is exhausting:
  "Filesystem Space in use: 908728/1011928 KB (90%)" !!

same as df (5% diference is the reserve for the superuser)

servidor:/cache# df
Filesystem 1k-blocks Used Available Use% Mounted on
/dev/hda8 1011928 908792 51732 95% /cache

It seems that squid is waiting for my 800Mb limit be reached. I don't
know why both accountings (squid and df) don't match since this partition is
exclusive for squid cache.

I have recreated (squid -z) it twice after runing out of space and don't want
to do it again. PLEASE! :-)

>
> Is the unlinkd process running?

servidor:/cache# ps ax |grep unlin
10937 ? SN 0:00 (unlinkd)

Always! I've checked that a long ago.

> >
> > I was invetigating a little and look what I found:
> >
> > servidor:/cache# du -ks *
> > 549196 00
> > 386644 01
> > 1028 02
> > 1028 03
>
> > Shouldn't it be an even distribution of files over the directory tree?
>
> no, see the FAQ.
>
> if you want them spread out more, decrease the number of L2 directories.

Thanks. Did not know that.

Any other information you need?

Ethy H. Brito /"\
InterNexo Ltda. \ / CAMPANHA DA FITA ASCII - CONTRA MAIL HTML
+(12) 341-6860 X ASCII RIBBON CAMPAIGN - AGAINST HTML MAIL
Sao Jose dos Campos / \

--
To unsubscribe, see http://www.squid-cache.org/mailing-lists.html