[squid-users] Re: squid multicast icp

OK, so it's not just me :-)

Using 2.4-STABLE7 I set up the conf file like this (from 10.40.33.12):

mcast_groups 239.255.40.10
cache_peer 239.255.40.10 multicast 3128 3130 ttl=4
cache_peer 10.40.33.10 sibling 3128 3130 proxy-only multicast-responder
cache_peer 10.40.33.11 sibling 3128 3130 proxy-only multicast-responder
cache_peer 10.40.33.13 sibling 3128 3130 proxy-only multicast-responder

I do a 'tcpdump -n -vv -x -X -s 150 host 10.40.33.10 and udp port icpv2'
and see the following:

14:26:11.588220 10.40.33.12.icpv2 > 10.40.33.10.icpv2: [udp sum ok] udp
100 (DF) (ttl 64, id 0, len 128)
0x0000 4500 0080 0000 4000 4011 e407 0a28 210c E.....@.@....(!.
0x0010 0a28 210a 0c3a 0c3a 006c 56d0 0302 0064 .(!..:.:.lV....d
0x0020 0000 1900 0000 0000 0000 0000 0000 0000 ................
0x0030 6874 7470 3a2f 2f36 352e 3231 342e 3336 http://65.214.36
0x0040 2e32 3430 2f65 3f74 3d70 6963 7475 7265 .240/e?t=picture
0x0050 732b 6f66 2b68 6f6d 6573 2b61 726f 756e s+of+homes+aroun
0x0060 642b 7468 652b 776f 726c 6426 663d 3026 d+the+world&f=0&
0x0070 733d 3026 613d 3026 6361 6368 653d 3000 s=0&a=0&cache=0.
14:26:15.051062 10.40.33.10.icpv2 > 10.40.33.12.icpv2: [udp sum ok] udp
98 (DF) (ttl 64, id 0, len 126)
0x0000 4500 007e 0000 4000 4011 e409 0a28 210a E..~..@.@....(!.
0x0010 0a28 210c 0c3a 0c3a 006a 5170 0302 0062 .(!..:.:.jQp...b
0x0020 0000 1a3a 0000 0000 0000 0000 0000 0000 ...:............
0x0030 6874 7470 3a2f 2f36 352e 3231 342e 3336 http://65.214.36
0x0040 2e32 3430 2f65 3f74 3d70 6963 7475 7265 .240/e?t=picture
0x0050 732b 6f6e 2b74 6865 2b6d 6f76 6965 2b74 s+on+the+movie+t
0x0060 6865 2b6d 6174 7269 7826 663d 3026 733d he+matrix&f=0&s=
0x0070 3026 613d 3026 6361 6368 653d 3000 0&a=0&cache=0.
14:26:17.458355 10.40.33.10.icpv2 > 10.40.33.12.icpv2: [udp sum ok] udp
42 (DF) (ttl 64, id 0, len 70)
0x0000 4500 0046 0000 4000 4011 e441 0a28 210a E..F..@.@..A.(!.
0x0010 0a28 210c 0c3a 0c3a 0032 d7cf 0302 002a .(!..:.:.2.....*
0x0020 0000 1a3b 0000 0000 0000 0000 0000 0000 ...;............
0x0030 6874 7470 3a2f 2f32 3339 2e32 3535 2e34 http://239.255.4
0x0040 302e 3130 2f00 0.10/.
14:26:22.626109 10.40.33.10.icpv2 > 239.255.40.10.icpv2: [udp sum ok] udp
89 (DF) (ttl 4, id 0, len 117)
0x0000 4500 0075 0000 4000 0411 333d 0a28 210a E..u..@...3=.(!.
0x0010 efff 280a 0c3a 0c3a 0061 cde2 0102 0059 ..(..:.:.a.....Y
0x0020 0000 1901 0000 0000 0000 0000 0000 0000 ................
0x0030 0000 0000 6874 7470 3a2f 2f36 352e 3231 ....http://65.21
0x0040 342e 3336 2e32 3430 2f65 3f74 3d6c 6173 4.36.240/e?t=las
0x0050 2b76 6567 6173 2b6d 6565 7469 6e67 2666 +vegas+meeting&f
0x0060 3d38 3026 733d 3830 2661 3d30 2663 6163 =80&s=80&a=0&cac
0x0070 6865 3d30 00 he=0.
14:26:22.626191 10.40.33.12.icpv2 > 10.40.33.10.icpv2: [udp sum ok] udp
85 (DF) (ttl 64, id 0, len 113)
0x0000 4500 0071 0000 4000 4011 e416 0a28 210c E..q..@.@....(!.
0x0010 0a28 210a 0c3a 0c3a 005d b8c4 0302 0055 .(!..:.:.].....U
0x0020 0000 1901 0000 0000 0000 0000 0000 0000 ................
0x0030 6874 7470 3a2f 2f36 352e 3231 342e 3336 http://65.214.36
0x0040 2e32 3430 2f65 3f74 3d6c 6173 2b76 6567 .240/e?t=las+veg
0x0050 6173 2b6d 6565 7469 6e67 2666 3d38 3026 as+meeting&f=80&
0x0060 733d 3830 2661 3d30 2663 6163 6865 3d30 s=80&a=0&cache=0
0x0070 00 .

This is very repeatable, the problem seems to be that the ICP packet at
14:26:15.051062 never gets responded to by squid. A longer capture with
less data looks like this:

14:33:30.592783 10.40.33.10.icpv2 > 239.255.40.10.icpv2: udp 94 (DF)
(ttl 4, id 0, len 122)
14:33:30.593035 10.40.33.12.icpv2 > 10.40.33.10.icpv2: udp 90 (DF)
(ttl 64, id 0, len 118)
14:33:41.068474 10.40.33.10.icpv2 > 10.40.33.12.icpv2: udp 98 (DF)
(ttl 64, id 0, len 126)
14:33:41.641678 10.40.33.10.icpv2 > 239.255.40.10.icpv2: udp 128 (DF)
(ttl 4, id 0, len 156)
14:33:41.641827 10.40.33.12.icpv2 > 10.40.33.10.icpv2: udp 124 (DF)
(ttl 64, id 0, len 152)
14:33:52.015601 10.40.33.10.icpv2 > 10.40.33.12.icpv2: udp 93 (DF)
(ttl 64, id 0, len 121)
14:33:52.722580 10.40.33.10.icpv2 > 239.255.40.10.icpv2: udp 118 (DF)
(ttl 4, id 0, len 146)
14:33:52.722838 10.40.33.12.icpv2 > 10.40.33.10.icpv2: udp 114 (DF)
(ttl 64, id 0, len 142)

As you can see, this pattern repeats itself, with .10 sending to multicast
and getting a reply, then sending out a regular ICP that doesn't get
replied to (and I'm not sure why .10 would send it, having .12 as a
multicast-responder).

Note that when squid first starts up, it does send out a multicast packet
for (apparently) each incoming query, then this stops about 15 sec after
it starts and the pattern above becomes the norm. I suspect squid may be
thinking the multicast group is down for some reason? The fact that .12
replies to the multicast packet (and pings work as expected to the
multicast group) makes me think it's not network related.

My squid is compiled as:
./configure --prefix=/usr/dh/squid --with-icp \
 '--enable-removal-policies=lru heap' --enable-snmp

I'm new to debugging squid so I'd greatly appriciate any suggestions that
might get more data to debug this problem!

I opened bug 439 for this with a copy of the data above...
                                                                -Rob

On Thu, 26 Sep 2002, fooler wrote:

> hi rob,
>
> there is one... adrian chadd told me if he have the time he will look
> into it... but unfortunately, upto the latest version of squid, multicast
> icp is still broken :-<
> our proxy farm is still using squid2.3stable4 up to now and cant migrate
> to the latest version because of this issue...
>
> by cc to squid mailing list and especially to the developers, any inputs
> about this? thanks.
>
> fooler.
>
>
> ----- Original Message -----
> From: "Rob Hagopian" <rob@hagopian.net>
> To: <fooler@skyinet.net>
> Sent: Thursday, September 26, 2002 3:55 AM
> Subject: squid multicast icp
>
>
> > Did you ever get a response to your post to squid-users (below)? I seem to
> > be having a similar problem...
> > -Rob
> >
> > Date: Sat, 5 May 2001 14:17:23 +0800
> > Reply-To: fooler <fooler@SKYINET.NET>
> > Sender: Fake list for archiving at Cineca <SQUID@LIST.CINECA.IT>
> > From: fooler <fooler@SKYINET.NET>
> > Subject: [squid-users] MULTICAST ICP BUGS ON SQUID 2.4STABLE1
> > Comments: To: squid-users@squid-cache.org
> > Content-Type: multipart/alternative;
> >
> > hi all,
> >
> > before, i was using squid 2.3stable4 using multicast icp and everything
> > works fine until i upgraded to squid 2.4stable1 with the same multicast
> > config before, cachemgr.cgi's peer cache statistics for squid 2.4stable1
> > shows that:
> >
> > Multicast G: 224.9.9.9/3128/3130 Flags : ttl=1 Address[0] : 224.9.9.9
> > Status : Down
> >
> > status is DOWN and the same time:
> >
> > Sibling : proxy.xxxxx.xxx/3128/3130 Flags : proxy-only multicast-responder
> > Address[0] : xxx.xxx.xxx.xxx.xxx Status : Up AVG RTT : 1 msec FETCHES : 0
> > 0% IGNORED : 487547 39% Histogram of PINGS ACKED: ICP_HIT : 479668 38%
> > ICP_MISS : 780841 62% ICP_MISS_NOF : 1955 0% keep-alive ratio: 98%
> >
> > FETCHES is 0% but ICP_HIT indicates that there are lots of hits. any fix
> > about this? thanks.
> >
> > fooler.
> >
>
Received on Thu Sep 26 2002 - 09:01:55 MDT