Some Eclipse Foundation services are deprecated, or will be soon. Please ensure you've read this important communication.

Bug 298021

Summary:	CVS timeouts to dev.eclipse.org
Product:	Community	Reporter:	Kim Moir <kim.moir>
Component:	CVS	Assignee:	Eclipse Webmaster <webmaster>
Status:	RESOLVED NOT_ECLIPSE	QA Contact:
Severity:	major
Priority:	P3	CC:	anne.jacko, daniel_megert, holmesc, karl.matthias, pwebster, remy.suen
Version:	unspecified
Target Milestone:	---
Hardware:	PC
OS:	Windows XP
Whiteboard:

Description Kim Moir

2009-12-16 18:14:05 EST

I'm seeing CVS timeouts during our build which is making our builds stall.  I can ssh without a problem from my build machine to dev.eclipse.org.  However, cvs is failing.  

-bash-3.00$  cvs -d kmoir@dev.eclipse.org:/cvsroot/eclipse export -d org.eclipse.core.net.win32.x86 -rI20080909 org.eclipse.core.net/fragments/org.eclipse.core.net.win32.x86
connect to address 206.191.52.50: Connection timed out

This is a repeatable problem.  Everytime it times out.

If I check out the same project from build.eclipse.org, there isn't a problem.

Comment 1 Denis Roy

2009-12-16 19:17:43 EST

Aren't you missing :ext: or :pserver: somewhere in there, or is it defined in an ENV variable?

Also, I'd need your internet-facing IP to see if there is a block.  Your Ottawa lab should be in the whitelist, but you guys bounce off routers in odd ways sometimes  :)

Comment 2 Kim Moir

2009-12-16 19:35:29 EST

no, the ext ssh thing is fine. Yes, we have an env variable.  I use ssh keys so it shouldn't be a problem.  We switched to using ssh on Dec 3.  Here is the firewall I used to login from the build machine.

fw54.torolab.ibm.com

Comment 3 Denis Roy

2009-12-16 19:51:32 EST

There's nothing on our end blocking your IP (it's whitelisted) and I can run the cvs export over SSH from my home connection.  Truly weird that you can SSH in from the machine.

Comment 4 Karl Matthias

2009-12-17 12:22:41 EST

Well some simple tracerouting from dev.eclipse.org to your address shows consistent packet loss between IBM's head-end router and the upstream that heads to us.  This is across two different paths at least.  I don't have this problem when tracerouting to the torolab from either my house or another site in Ohio.  I would ask your network people if they know anything about it.

Comment 5 Kim Moir

2009-12-17 16:13:24 EST

Thanks Karl.  I've opened a ticket with them to investigate.

Comment 6 Craig Holmes

2009-12-18 11:32:39 EST

Can you please provide the traceroute?

Comment 7 Karl Matthias

2009-12-18 12:09:54 EST

Looks fine this morning.  Yesterday the failures looked like they were between IBM and Allstream.  Sorry I didn't post the traceroutes.  I no longer have them.

Comment 8 Craig Holmes

2009-12-18 12:17:40 EST

Interesting, I wasn't aware that eclipse traffic was flowing via our allstream link. I'll investigate that further.

Kim - Our WAN link at QV is pinned right now, but perhaps a little later in the day you could try CVS again.

Comment 9 Denis Roy

2009-12-21 13:39:25 EST

> Kim - Our WAN link at QV is pinned right now


Sounds like someone's pipe is fuller than ours  =)