Marcelus Trojahn wrote:
Hello,
For a few weeks now I've been tuning a squid to my network. I have
around 80Mb/s of HTTP traffic going through this server during most
of the day. I had a little of every possible trouble I can think of
and most of them are solved now after a few more white hair in my head
:)
Anyway... My only problem right now looks like some sort of disk I/O
problem. Consider the following:
File descriptor usage for squid:
Maximum number of file descriptors: 32768
Largest file desc currently in use: 24143
Number of file desc currently in use: 23257
Files queued for open: 2
Available number of file descriptors: 9509
Reserved number of file descriptors: 100
Store Disk files open: 11399
Squid isn't really slow or anything right now... But the value on the
"Store disk files open" apparently never goes down and it is starting
to concern me...
Is there a way for some of this opened files to be stale, maybe?
Should I start to worry about a incoming crash or I just need to
recompile the whole thing with more max FDs?
Here's a little more info just to show the current load:
Squid Object Cache: Version 3.1.0.13-20090807
Start Time: Fri, 21 Aug 2009 21:53:42 GMT
Current Time: Sat, 22 Aug 2009 02:10:12 GMT
Connection information for squid:
Number of clients accessing cache: 1714
Number of HTTP requests received: 6344801
Number of ICP messages received: 0
Number of ICP messages sent: 0
Number of queued ICP replies: 0
Number of HTCP messages received: 0
Number of HTCP messages sent: 0
Request failure ratio: 0.00
Average HTTP requests per minute since start: 24736.6
Average ICP messages per minute since start: 0.0
Select loop called: 189242493 times, 0.081 ms avg
Cache information for squid:
Hits as % of all requests: 5min: 31.9%, 60min: 36.2%
Hits as % of bytes sent: 5min: 8.1%, 60min: 8.9%
Memory hits as % of hit requests: 5min: 27.2%, 60min: 26.7%
Disk hits as % of hit requests: 5min: 39.9%, 60min: 37.5%
Storage Swap size: 97730572 KB
Storage Swap capacity: 95.4% used, 4.6% free
Storage Mem size: 4096128 KB
Storage Mem capacity: 100.0% used, 1801439850948198.5% free
Mean Object Size: 29.26 KB
Requests given to unlinkd: 0
Median Service Times (seconds) 5 min 60 min:
HTTP Requests (All): 0.16775 0.23230
Cache Misses: 0.23230 0.30459
Cache Hits: 0.00091 0.02190
Near Hits: 0.12106 0.25890
Not-Modified Replies: 0.00000 0.00000
DNS Lookups: 0.00190 0.01153
ICP Queries: 0.00000 0.00000
Resource usage for squid:
UP Time: 15389.656 seconds
CPU Time: 6403.690 seconds
CPU Usage: 41.61%
CPU Usage, 5 minute avg: 36.83%
CPU Usage, 60 minute avg: 41.76%
Process Data Segment Size via sbrk(): 5909608 KB
Maximum Resident Size: 0 KB
Page faults with physical i/o: 8560
The cache filesystem, after a bit of test, was chosen to be ext2 and
squid was compiled like this:
./configure --prefix=/usr --build=x86_64-pc-linux-gnu
--host=x86_64-pc-linux-gnu --mandir=/usr/share/man
--infodir=/usr/share/info --datadir=/usr/share --sysconfdir=/etc
--localstatedir=/var/lib --libdir=/usr/lib64 --sysconfdir=/etc/squid
--libexecdir=/usr/libexec/squid --localstatedir=/var
--datadir=/usr/share/squid --with-logdir=/var/log/squid
--with-default-user=squid --enable-removal-policies=lru,heap
--enable-external-acl-helpers=ip_user,session,unix_group
--enable-useragent-log --disable-cache-digests --with-large-files
--with-filedescriptors=32768 --disable-snmp --disable-ssl
--disable-icap-client --enable-zph-qos --disable-ipv6 --enable-caps
--enable-linux-netfilter --enable-async-io=36
--
Marcelus Trojahn
The easy way to tell is checking them all at:
squidclient mgr:filedescriptors
squidclient mgr:openfd_objects
Which give you a list of a) what every FD tracked by Squid is doing and
b) what the ones that apply to disk objects are doing.
[WARNING: detailed explanation of FD usage follows, you might not want
to know]
Your stats say this:
Number of clients accessing cache: 1714
+
Store Disk files open: 11399
Some of the files might be stale, yes.
However the FD accounted include TCP/UDP sockets etc.
Each of the 1714 clients is using:
1 for the link to the client
1 for a disk/memory-backed object they are receiving/sending
1 for the link to the server it came from (TCP_MISSes only)
So up to:
1714 + 1714 + 1714 = 5142 FD will be in use for regular traffic.
Add to that:
the number of helper you have running.
the number of *_port lines you have in squid.conf
the number of DNS nameservers in use (for TCP DNS)
the number of DNS requests made via UDP in the last 30 seconds or
so (your system UDP timeout tuning alters this) less those which have an
answer returned _multiplied_ by the number of nameservers in use.
the number of files unlinkd is currently queued waiting for
erasure/RELEASE.
Amos
--
Please be using
Current Stable Squid 2.7.STABLE6 or 3.0.STABLE18
Current Beta Squid 3.1.0.13
