[squid-users] Possible filedescriptor leak?

I have just implemented the bugfix for aufs by building and installing
the snapshot of 20021112. I am now seeing a strange symptom in file
descriptor usage, where "Files queued for open is monotonically
increasing.

Squid -v output is:
Squid Cache: Version 2.5.STABLE1-20021112
configure options: --enable-dlmalloc --enable-gnuregex
--enable-xmalloc-statistics --enable-async-io=160 --enable-useragent-log
--enable-referer-log --disable-wccp --enable-snmp --enable-poll
--enable-err-languages=English '--enable-removal-policies=lru heap'

Here is some data from today's production squid...

At 0900:

File descriptor usage for squid:
        Maximum number of file descriptors: 1024
        Largest file desc currently in use: 45
        Number of file desc currently in use: 25
        Files queued for open: 124
        Available number of file descriptors: 875
        Reserved number of file descriptors: 100
        Store Disk files open: 1

At 1000:

File descriptor usage for squid:
        Maximum number of file descriptors: 1024
        Largest file desc currently in use: 50
        Number of file desc currently in use: 36
        Files queued for open: 127
        Available number of file descriptors: 861
        Reserved number of file descriptors: 100
        Store Disk files open: 0

At 1100:

File descriptor usage for squid:
        Maximum number of file descriptors: 1024
        Largest file desc currently in use: 46
        Number of file desc currently in use: 43
        Files queued for open: 133
        Available number of file descriptors: 848
        Reserved number of file descriptors: 100
        Store Disk files open: 1

As you can see, it is not a particularly busy cache with only a small
number of concurrent active connections. The Process Filedescriptor
Allocation page, AsyncIO Function Counters page and others show nothing
out of the ordinary (AFAICT). I cannot vouch for the accuracy of my
failing memory but the previous snapshot I used (20021028) didn't show
the same symptom AFAIK (it ran for 15 days.)

Anyone else seen this? Is it perhaps a simple reporting problem? Is it
a new feature where a file cache is being maintained? ;-)

Michael Lightfoot
Unix Consultant
ISG Host Systems
Comcare
+61 2 62750680
Apologies for the rubbish that follows...
------------------------------------------------------------------------
NOTICE: This e-mail message and attachments may contain confidential
information. If you are not the intended recipient you should not use or
disclose any information in the message or attachments. If received in
error, please notify the sender by return email immediately. Comcare
does not waive any confidentiality or privilege.

Comcare's Canberra and Melbourne offices have moved to new premises.

We hope this notice makes it easier for you to update your records.
 
New street addresses:
 
Canberra office
Level 1
14 Moore street
CANBERRA CITY ACT 2601

Melbourne Office
Level 2
121 William Street
MELBOURNE VIC 3001
Claims Fax Number 03 9620 4721

Please call Comcare on 1300 366 979 if you have any queries.
Received on Wed Nov 13 2002 - 17:10:26 MST