You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@trafficserver.apache.org by "Miles Libbey (JIRA)" <ji...@apache.org> on 2010/04/19 23:02:50 UTC
[jira] Commented: (TS-32) Fix ICP
[ https://issues.apache.org/jira/browse/TS-32?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12858681#action_12858681 ]
Miles Libbey commented on TS-32:
--------------------------------
(was yahoo bug 1999704 and 3140708)
-----Original Message-----
From: Ethan Lai
Sent: Friday, June 13, 2008 4:41 AM
Subject: [Cannot enable ICP on YTS 1.17.x
Hi TS,
We're trying to build a TS cluster serving for same small contents.
While running TS 1.16.[2|4], it works fine if we set icp.enabled INT 2
and have some icp Sibling in icp.config.
But it core dumps while starting up TS 1.17.[0|7] with same
configuration files.
Here is some info:
icp.config:
p1.cover.wretch.tp2.yahoo.net::2:80:3130:0:0.0.0.0:1:
p2.cover.wretch.tp2.yahoo.net::2:80:3130:0:0.0.0.0:1:
traffic.out:
[Jun 13 19:14:59.569] {4157241024} STATUS: opened
/home/y/logs/yts/diags.log
Inside Connection::bind_connect in UnixNet.cc
Inside Connection::bind_connect in UnixNet.cc
NOTE: Traffic Server received Sig 11: Segmentation fault
bin/traffic_server STACK TRACE:
0xFFFFE500::__kernel_sigreturn
0x08326B5B:bin/traffic_server:_ZN15UDPNetProcessor11recvfrom_reEP12ContinuationPviP8sockaddrPiP13IOBufferBlockibi
0xA53F8000:[0xa53f8000]
0x080E3C4C:bin/traffic_server:_ZN15ICPPeerReadCont16ICPPeerReadEventEiP5Event
0x080E44E5:bin/traffic_server:_ZN14ICPHandlerCont13PeriodicEventEiP5Event
0x08338CB2:bin/traffic_server:_ZN7EThread13process_eventEP5Eventi
0x0833954E:bin/traffic_server:_ZN7EThread7executeEv
0x08137903:bin/traffic_server:main
0x006B5DE3:/lib/tls/libc.so.6:__libc_start_main
0x080D4B61:bin/traffic_server:__gxx_personality_v0
0x080D4B61:bin/traffic_server:__gxx_personality_v0
0x080D4B61:bin/traffic_server:__gxx_personality_v0
0x080D4B61:bin/traffic_server:__gxx_personality_v0
gdb output:
#0 0x08326b5b in UDPNetProcessor::recvfrom_re (this=0x853d488,
cont=0x9dbb07c8, token=0x9dbb07c8, fd=31,
fromaddr=0xa5cb0cb4, fromaddrlen=0xa5cb0cc4, buf=0x9dbb8ff8,
len=16347, useReadCont=true, timeout=0)
at ../../../iocore/net/P_UnixPollDescriptor.h:28
#1 0x080e8906 in ParentSiblingPeer::RecvFrom_re (this=0xbf04f030,
cont=0x9dbb07c8, token=0x9dbb07c8, bufblock=0x9dbb8ff8,
size=16347, from=0xa5cb0cb4, fromlen=0xa5cb0cc4) at
../../proxy/ICPConfig.cc:938
#2 0x080e207b in ICPPeerReadCont::PeerReadStateMachine
(this=0x9dbb07c8, s=0x9dbb7f20, e=0x0)
at ../../libinktomi++/Ptr.h:213
#3 0x080e3c4c in ICPPeerReadCont::ICPPeerReadEvent
(this=0x9dbb07c8, event=2, e=0x0) at ../../proxy/ICP.cc:361
#4 0x080e44e5 in ICPHandlerCont::PeriodicEvent (this=0xa5cb0f48,
event=2, e=0x8612fd0)
at ../../iocore/eventsystem/I_Continuation.h:103
#5 0x08338cb2 in EThread::process_event (this=0xbf04d008,
e=0x8612fd0, calling_code=2)
at ../../../iocore/eventsystem/I_Continuation.h:103
#6 0x0833954e in EThread::execute (this=0xbf04d008) at
../../../iocore/eventsystem/UnixEThread.cc:188
#7 0x08137903 in main (argc=3, argv=0xffffd614) at
../../iocore/eventsystem/P_Thread.h:21
Do you team have any ideas?
Please advise and many thanks,
Ethan
Original description
by Ravi Teja 5 months ago at 2009-10-26 05:49
When I enable ICP in trafficserver-1.17.2, TS is dumping core and not starting at all. But, when I try with same
configuration in version trafficserver-1.16.2, TS starts without any problems.
----icp configuration----
yqlyts1.pipes.mud.yahoo.com:209.191.71.34:2:80:3130:0:0.0.0.0:1:
yqlyts2.pipes.mud.yahoo.com:209.191.71.35:2:80:3130:0:0.0.0.0:1:
----logs ----
[Oct 26 04:50:26.669] Manager {4157041024} ERROR: [LocalManager::pollMgmtProcessServer] Server Process
terminated due to Sig 11: Segmentation fault
[Oct 26 04:50:26.669] Manager {4157041024} ERROR: (last system error 9: Bad file descriptor)
[Oct 26 04:50:26.669] Manager {4157041024} ERROR: [Alarms::signalAlarm] Server Process was reset
[Oct 26 04:50:26.670] Manager {4157041024} ERROR: (last system error 9: Bad file descriptor)
[example_alarm_bin.sh] sent alarm: yqlyts2.pipes.mud.yahoo.com [TrafficManager] Traffic Server process
was reset.
[example_prep.sh] Checking/Moving old cores...
[Oct 26 04:50:29.322] {4150838976} STATUS: opened /home/y/logs/yts/diags.log
NOTE: Traffic Server received Sig 11: Segmentation fault
bin/traffic_server STACK TRACE:
0xFFFFE500::__kernel_sigreturn
0x08339F6B:bin/traffic_server:_ZN15UDPNetProcessor11recvfrom_reEP12ContinuationPviP8sockaddrPiP13IOBuf
ferBlockibi
0x00004000:[0x4000]
0x080E5FDC:bin/traffic_server:_ZN15ICPPeerReadCont16ICPPeerReadEventEiP5Event
0x080E6875:bin/traffic_server:_ZN14ICPHandlerCont13PeriodicEventEiP5Event
0x0834CBE2:bin/traffic_server:_ZN7EThread13process_eventEP5Eventi
0x0834D47E:bin/traffic_server:_ZN7EThread7executeEv
0x0834A65C:bin/traffic_server:(null)
0xF7F373CC:/lib/tls/libpthread.so.0:(null)
0x00B15F0E:/lib/tls/libc.so.6:__clone
0x00B15F0E:/lib/tls/libc.so.6:__clone
0x00B15F0E:/lib/tls/libc.so.6:__clone
0x00B15F0E:/lib/tls/libc.so.6:__clone
Comment 1
by Bryan Call 5 months ago at 2009-10-26 08:42:10
Yes, this is a know issue in the 1.17.x branch. ICP hasn't been
working since we reworked how we handle socket connections.
-Bryan
> Fix ICP
> -------
>
> Key: TS-32
> URL: https://issues.apache.org/jira/browse/TS-32
> Project: Traffic Server
> Issue Type: Bug
> Components: Core
> Reporter: Miles Libbey
>
> http://icp.ircache.net/
> The ICP implementation in Traffic Server broke when epoll() was introduced. Its still an interesting and used feature in caches:
> - when a caching layer of several boxes are used ICP helps to reduce disparities when a client is not routed to the same cache on subsequent requests
> - after a restart, it can help reduce the time spent in a cold cache situation
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.